)]}'
{"specs/queens/oslo-config-db.rst":[{"author":{"_account_id":1179,"name":"Clay Gerrard","email":"clay.gerrard@gmail.com","username":"clay-gerrard"},"change_message_id":"d7ac8fb1537cae2df504fd444778fd233c7e997d","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Currently, oslo.config is tightly bound to the plain text"},{"line_number":8,"context_line":"configuration files. This spec describes changes to allow deployers to"},{"line_number":9,"context_line":"store configuration data in a Configuration Management Database (CMDB)"},{"line_number":10,"context_line":"such as etcd."},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_b0f008ff","line":10,"updated":"2017-04-08 18:38:36.000000000","message":"It\u0027s not really clear to me that etcd meets the classical definition of a CMDB [1]?  Perhaps the definition has evolved.  Does etcd normally keep an operator accessible \"history\" of who made what change when and why? \n\nhttps://en.wikipedia.org/wiki/Configuration_management_database","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"5df920c8a26e8305661b866501863330896b7c1f","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Currently, oslo.config is tightly bound to the plain text"},{"line_number":8,"context_line":"configuration files. This spec describes changes to allow deployers to"},{"line_number":9,"context_line":"store configuration data in a Configuration Management Database (CMDB)"},{"line_number":10,"context_line":"such as etcd."},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_18126f0a","line":10,"in_reply_to":"bff0334d_9f106faa","updated":"2017-04-18 12:11:06.000000000","message":"+1 for the new wording, Doug.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":6928,"name":"Ben Nemec","email":"openstack@nemebean.com","username":"bnemec"},"change_message_id":"f515603e8005fbc876e3bcce9f417779857d0a11","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Currently, oslo.config is tightly bound to the plain text"},{"line_number":8,"context_line":"configuration files. This spec describes changes to allow deployers to"},{"line_number":9,"context_line":"store configuration data in a Configuration Management Database (CMDB)"},{"line_number":10,"context_line":"such as etcd."},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_6344341c","line":10,"in_reply_to":"bff0334d_9f106faa","updated":"2017-05-05 19:43:29.000000000","message":"Maybe Configuration Storage Database if it doesn\u0027t provide the history that a Management Database would?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"830ccc3e71773ed798b1909e3974b7d97360c28e","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Currently, oslo.config is tightly bound to the plain text"},{"line_number":8,"context_line":"configuration files. This spec describes changes to allow deployers to"},{"line_number":9,"context_line":"store configuration data in a Configuration Management Database (CMDB)"},{"line_number":10,"context_line":"such as etcd."},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_9f106faa","line":10,"in_reply_to":"bff0334d_b0f008ff","updated":"2017-04-09 19:02:10.000000000","message":"Excellent question. I kept some of the original text from the first versions of this proposal, and the term CMDB may not be the best choice here.\n\nWhat would you propose as an alternative instead of \"a central service where the configuration is stored,\" which is a bit of a mouthful?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":1179,"name":"Clay Gerrard","email":"clay.gerrard@gmail.com","username":"clay-gerrard"},"change_message_id":"d7ac8fb1537cae2df504fd444778fd233c7e997d","unresolved":false,"context_lines":[{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Centralized configuration management is a highly desired feature as it"},{"line_number":16,"context_line":"allows having a single point of control of all the configuration"},{"line_number":17,"context_line":"parameters for OpenStack components. However, the existing"},{"line_number":18,"context_line":"implementation of oslo.config does not support an option to retrieve"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_10ba1409","line":15,"updated":"2017-04-08 18:38:36.000000000","message":"It may not be appropriate to go into more detail in this spec - but for those uneducated in the highly desired feature(s) of etcd for configuration [1] do you have any reference material on hand you might link - either in a comment on this spec or as a permanent link in references?  Is there a good synopsis in the ML thread?\n\nI clicked through a few eatherpads hoping to find \"use-case\" [2] but most just jump straight into \"how to implement\" instead of why - which makes sense - the folks most interested in doing the work already understand why the work is valuable.  But... assuming buy-in ,pulling in the same direction, vision etc is useful\n\n    1. o/\n    2. https://etherpad.openstack.org/p/tripleo-etcd-transition seems to say:\n           deploying config files to containers is a bad architecture (!?)","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"830ccc3e71773ed798b1909e3974b7d97360c28e","unresolved":false,"context_lines":[{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Centralized configuration management is a highly desired feature as it"},{"line_number":16,"context_line":"allows having a single point of control of all the configuration"},{"line_number":17,"context_line":"parameters for OpenStack components. However, the existing"},{"line_number":18,"context_line":"implementation of oslo.config does not support an option to retrieve"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_5f266717","line":15,"in_reply_to":"bff0334d_10ba1409","updated":"2017-04-09 19:02:10.000000000","message":"As I understand it, generating a container with a config file is less appealing than generating a container in which a single value (the URL for retrieving the config) can be fetched. The container images are more reusable in the latter case.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"e251a7c2b24436f30cd1c346f384c8546a7dcdab","unresolved":false,"context_lines":[{"line_number":12,"context_line":"Problem description"},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Centralized configuration management is a highly desired feature as it"},{"line_number":16,"context_line":"allows having a single point of control of all the configuration"},{"line_number":17,"context_line":"parameters for OpenStack components. However, the existing"},{"line_number":18,"context_line":"implementation of oslo.config does not support an option to retrieve"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_815ecd41","line":15,"in_reply_to":"bff0334d_5f266717","updated":"2017-04-10 09:07:31.000000000","message":"What Doug said + some other aspects :\n\n* security. Instead of storing sensitive data (IP, credentials, etc), these values are store in a secure datastore (using TLS for data transport and authentication for accessing it). Which is much better than a configuration file in /etc\n\n* Migrating containers becomes easier (in a kubernetes environment for example), where a container can be moved around  nodes without managing the configuration files.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":7385,"name":"Thomas Herve","email":"therve@redhat.com","username":"therve"},"change_message_id":"e8bbcd0f6cdf34f178a421bbf49220de3d6b4ab8","unresolved":false,"context_lines":[{"line_number":28,"context_line":""},{"line_number":29,"context_line":"The ``--config-file`` command line option is changed to"},{"line_number":30,"context_line":"``--config-url``, with the old option name retained as a deprecated"},{"line_number":31,"context_line":"version."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The scheme portion of the URL is used to determine which driver to"},{"line_number":34,"context_line":"use. The URL components other than the scheme are left up to the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_901064aa","line":31,"updated":"2017-04-08 19:29:59.000000000","message":"Is there much value to do that? Considering we need to keep config-dir, I\u0027m not sure it helps much except creating a lot of work for no functionality.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"830ccc3e71773ed798b1909e3974b7d97360c28e","unresolved":false,"context_lines":[{"line_number":28,"context_line":""},{"line_number":29,"context_line":"The ``--config-file`` command line option is changed to"},{"line_number":30,"context_line":"``--config-url``, with the old option name retained as a deprecated"},{"line_number":31,"context_line":"version."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The scheme portion of the URL is used to determine which driver to"},{"line_number":34,"context_line":"use. The URL components other than the scheme are left up to the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_3f1f23b8","line":31,"in_reply_to":"bff0334d_901064aa","updated":"2017-04-09 19:02:10.000000000","message":"It makes the name more accurate, but that was the only real reason to change. It\u0027s not that expensive in terms of the work to do in the library itself.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9796,"name":"ChangBo Guo","email":"glongwave@gmail.com","username":"gcb"},"change_message_id":"caa7c0b113db848578acfc331b8509dfb40cdacf","unresolved":false,"context_lines":[{"line_number":43,"context_line":"Each occurance of ``--config-url`` adds a new \"source\" to the end of a"},{"line_number":44,"context_line":"search path. When ConfigOpt looks for an option, it goes through the"},{"line_number":45,"context_line":"sources in the order they are provided, starting with the command line"},{"line_number":46,"context_line":"options and then the sources provided by the user.  The first source"},{"line_number":47,"context_line":"that provides a configured value for an option causes the search to"},{"line_number":48,"context_line":"end (sources are not \"merged\")."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"An occurance of ``--config-dir`` will continue to be treated as a"},{"line_number":51,"context_line":"single source, as it is now (all of the options are merged into one"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_c400c037","line":48,"range":{"start_line":46,"start_character":52,"end_line":48,"end_character":31},"updated":"2017-04-17 12:49:52.000000000","message":"Currentlater one wins see https://github.com/openstack/oslo.config/blob/master/oslo_config/cfg.py#L167-L170,\n\nMaybe we need document the change.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":6928,"name":"Ben Nemec","email":"openstack@nemebean.com","username":"bnemec"},"change_message_id":"f515603e8005fbc876e3bcce9f417779857d0a11","unresolved":false,"context_lines":[{"line_number":43,"context_line":"Each occurance of ``--config-url`` adds a new \"source\" to the end of a"},{"line_number":44,"context_line":"search path. When ConfigOpt looks for an option, it goes through the"},{"line_number":45,"context_line":"sources in the order they are provided, starting with the command line"},{"line_number":46,"context_line":"options and then the sources provided by the user.  The first source"},{"line_number":47,"context_line":"that provides a configured value for an option causes the search to"},{"line_number":48,"context_line":"end (sources are not \"merged\")."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"An occurance of ``--config-dir`` will continue to be treated as a"},{"line_number":51,"context_line":"single source, as it is now (all of the options are merged into one"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_e94a65aa","line":48,"range":{"start_line":46,"start_character":52,"end_line":48,"end_character":31},"in_reply_to":"7ffa3b31_3c4bf12b","updated":"2017-05-05 19:43:29.000000000","message":"+1 to keeping the existing behavior.  Mostly because that\u0027s how the stuff I work with regularly works so it\u0027s less confusing for me. :-)","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"dc7277342878da6bf260081999c2edab5467bc51","unresolved":false,"context_lines":[{"line_number":43,"context_line":"Each occurance of ``--config-url`` adds a new \"source\" to the end of a"},{"line_number":44,"context_line":"search path. When ConfigOpt looks for an option, it goes through the"},{"line_number":45,"context_line":"sources in the order they are provided, starting with the command line"},{"line_number":46,"context_line":"options and then the sources provided by the user.  The first source"},{"line_number":47,"context_line":"that provides a configured value for an option causes the search to"},{"line_number":48,"context_line":"end (sources are not \"merged\")."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"An occurance of ``--config-dir`` will continue to be treated as a"},{"line_number":51,"context_line":"single source, as it is now (all of the options are merged into one"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_3c4bf12b","line":48,"range":{"start_line":46,"start_character":52,"end_line":48,"end_character":31},"in_reply_to":"7ffa3b31_c400c037","updated":"2017-04-17 21:22:13.000000000","message":"Interesting, and a very good catch. We could just build the stack in the other order instead to maintain the current behavior.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":1179,"name":"Clay Gerrard","email":"clay.gerrard@gmail.com","username":"clay-gerrard"},"change_message_id":"d7ac8fb1537cae2df504fd444778fd233c7e997d","unresolved":false,"context_lines":[{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_5049bc2e","line":65,"updated":"2017-04-08 18:38:36.000000000","message":"OH!  this is perhaps very valuable to understand?  The use-case is only for containerized deployments - people *do* in-fact expect to deployers will continue to deploy many OpenStack services via plain old config files - and not just because we\u0027re luddites - but because the value of the feature only makes sense in the context of containerized deployments.\n\nIs that true?\n\nIt seems like TripleO wants the services themselves in the containers to use the central k/v store - why is good for one type not equally good for another type?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":14985,"name":"Alex Schultz","email":"aschultz@next-development.com","username":"mwhahaha"},"change_message_id":"e72060a00c5f8323e6b9e684f32043c765d91902","unresolved":false,"context_lines":[{"line_number":61,"context_line":"deployments are most likely to pass one or more URLs to the container"},{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_ec04544c","line":65,"range":{"start_line":64,"start_character":7,"end_line":65,"end_character":23},"updated":"2017-04-10 22:32:40.000000000","message":"So I think it would beneficial to use both, but the question comes in how do you set priority of the sources. For example if I\u0027m deploying containers with some pre-baked in configurations but would like to include the single CMDB source as a way to apply overrides to the backed in configurations.  It would also help with the migration from config files into CMDB if you can continue to load from one and migrate over to the other over a longer period of time.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"5df920c8a26e8305661b866501863330896b7c1f","unresolved":false,"context_lines":[{"line_number":61,"context_line":"deployments are most likely to pass one or more URLs to the container"},{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_f8cb4350","line":65,"range":{"start_line":64,"start_character":7,"end_line":65,"end_character":23},"in_reply_to":"7ffa3b31_fc0229b4","updated":"2017-04-18 12:11:06.000000000","message":"I agree with Alex \u0026 Kevin, mixing might be an interesting feature for deployers. Though I would like to know why storing secrets is not the right solution Kevin? Do you think it\u0027s better to have them in a config file?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2638fe20ab418410528aadf8528ebb0b66004626","unresolved":false,"context_lines":[{"line_number":61,"context_line":"deployments are most likely to pass one or more URLs to the container"},{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_fc0229b4","line":65,"range":{"start_line":64,"start_character":7,"end_line":65,"end_character":23},"in_reply_to":"9ff52f3f_415152d8","updated":"2017-04-17 21:45:00.000000000","message":"I suppose we could remove that caveat. The only real issue I see is that it makes it more important for a person debugging the config to understand the search behavior. We could address that with a log message after all data sources are loaded, I guess.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"830ccc3e71773ed798b1909e3974b7d97360c28e","unresolved":false,"context_lines":[{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_1f1c5fbf","line":65,"in_reply_to":"bff0334d_5049bc2e","updated":"2017-04-09 19:02:10.000000000","message":"The feature makes container deployment easier. Whether it has any benefit for non-container deployments is not really something I considered. There\u0027s no particular reason not to do it, but I\u0027m also not sure if you already have a system for pushing config files out to servers that it makes a lot of sense to replace that just for the sake of doing so.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"e81f60dcadf39526657c90f644d72bf019190454","unresolved":false,"context_lines":[{"line_number":61,"context_line":"deployments are most likely to pass one or more URLs to the container"},{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9ff52f3f_415152d8","line":65,"range":{"start_line":64,"start_character":7,"end_line":65,"end_character":23},"in_reply_to":"bff0334d_57b28e4a","updated":"2017-04-13 22:48:09.000000000","message":"I think mixing might actually be pretty important for some cases. centralizing config can be very useful. storing secrets there may not be the right solution though.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"194b5e8620132ba31b0676bff5b7a2721c0ff618","unresolved":false,"context_lines":[{"line_number":61,"context_line":"deployments are most likely to pass one or more URLs to the container"},{"line_number":62,"context_line":"and use only the remote CMDB. Non-containerized deployments are most"},{"line_number":63,"context_line":"likely to pass a single file or use ``--config-dir`` to load multiple"},{"line_number":64,"context_line":"files. While the implementation would allow mixing files and CMDB,"},{"line_number":65,"context_line":"that is not recommended."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Implementation details"},{"line_number":68,"context_line":"----------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_57b28e4a","line":65,"range":{"start_line":64,"start_character":7,"end_line":65,"end_character":23},"in_reply_to":"bff0334d_ec04544c","updated":"2017-04-11 14:49:11.000000000","message":"See line 43. We\u0027ll need to make that search path behavior more clear in the documentation.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"c45140e996ce36c16cf5e5847f74fb658b5a345b","unresolved":false,"context_lines":[{"line_number":117,"context_line":"same syntax that the INI files use."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"If the value is not present in the database, the driver should raise"},{"line_number":120,"context_line":"NoValue. We can\u0027t use None as a sentinal indicating a missing value,"},{"line_number":121,"context_line":"so we use a custom exception type instead."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"The opt parameter is provided in case the driver needs to do advanced"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_0dc8c81f","line":120,"updated":"2017-04-07 23:00:08.000000000","message":"Maybe we should have it return a NoValue instance instead, and save exceptions for true errors?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"5d0097314df94034d8ffa2284fa1fb4e82fee06b","unresolved":false,"context_lines":[{"line_number":117,"context_line":"same syntax that the INI files use."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"If the value is not present in the database, the driver should raise"},{"line_number":120,"context_line":"NoValue. We can\u0027t use None as a sentinal indicating a missing value,"},{"line_number":121,"context_line":"so we use a custom exception type instead."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"The opt parameter is provided in case the driver needs to do advanced"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_7e1166ac","line":120,"in_reply_to":"3f044301_497719d6","updated":"2017-05-05 20:56:04.000000000","message":"Yeah, I was mostly thinking out loud there. The driver shouldn\u0027t know about the Opt default, because there may be multiple sources and the driver doesn\u0027t know it\u0027s position in the hierarchy.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":6928,"name":"Ben Nemec","email":"openstack@nemebean.com","username":"bnemec"},"change_message_id":"f515603e8005fbc876e3bcce9f417779857d0a11","unresolved":false,"context_lines":[{"line_number":117,"context_line":"same syntax that the INI files use."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"If the value is not present in the database, the driver should raise"},{"line_number":120,"context_line":"NoValue. We can\u0027t use None as a sentinal indicating a missing value,"},{"line_number":121,"context_line":"so we use a custom exception type instead."},{"line_number":122,"context_line":""},{"line_number":123,"context_line":"The opt parameter is provided in case the driver needs to do advanced"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_497719d6","line":120,"in_reply_to":"bff0334d_0dc8c81f","updated":"2017-05-05 19:43:29.000000000","message":"If the value is not present, but the opt exists, shouldn\u0027t it return the default for the opt?\n\nOh, but we don\u0027t want the driver to report a value when one isn\u0027t set because that might override a different source.  Since this would all be hidden internally to oslo.config I don\u0027t have strong feelings one way or another.  I\u0027d be fine deferring this question until implementation time and doing whatever is cleaner.  We\u0027ll need to define the behavior at some point, but I\u0027m not sure that has to be now.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":10787,"name":"Michal Jastrzebski (inc0)","email":"michal.jastrzebski@intel.com","username":"inc0"},"change_message_id":"b1edce584982e87fda669b2234246afa06857547","unresolved":false,"context_lines":[{"line_number":134,"context_line":"used in similar situations by the drivers."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"A ConfigurationSource subclass for managing the command line options"},{"line_number":137,"context_line":"will be created so that option lookup can be standardized."},{"line_number":138,"context_line":""},{"line_number":139,"context_line":"Drivers will be implemented inside the oslo.config library, and loaded"},{"line_number":140,"context_line":"through entry points managed by stevedore_ using the namespace"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_550bca7b","line":137,"updated":"2017-04-13 02:43:43.000000000","message":"One big issue I can see is organization of configs in kv storage itself. Per-host configuration mgmt to be precise. I\u0027d suggest doing something like this:\nconfig.nova.nova-compute.hostname.group_name.option_name takes precedence over\nconfig.nova.nova-compute.group.opt ... which takes precedence over config.nova.group.conf","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2638fe20ab418410528aadf8528ebb0b66004626","unresolved":false,"context_lines":[{"line_number":134,"context_line":"used in similar situations by the drivers."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"A ConfigurationSource subclass for managing the command line options"},{"line_number":137,"context_line":"will be created so that option lookup can be standardized."},{"line_number":138,"context_line":""},{"line_number":139,"context_line":"Drivers will be implemented inside the oslo.config library, and loaded"},{"line_number":140,"context_line":"through entry points managed by stevedore_ using the namespace"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_3c953141","line":137,"in_reply_to":"7ffa3b31_52838405","updated":"2017-04-17 21:45:00.000000000","message":"The spec is written in such a way that any host-specific portion of the URL should come before the app, group, field, etc. That way the deployer can build whatever host-specific scheme they want, using whatever level of complexity they like, and we\u0027ll just use it. For example, see the unique-client-identifier portion of the URL example in the etcd driver discussion starting on line 249.\n\nWith repeated instances of the --config-url option, it would be possible to have all three of global, host group, and individual host URLs using different prefixes.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"e81f60dcadf39526657c90f644d72bf019190454","unresolved":false,"context_lines":[{"line_number":134,"context_line":"used in similar situations by the drivers."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"A ConfigurationSource subclass for managing the command line options"},{"line_number":137,"context_line":"will be created so that option lookup can be standardized."},{"line_number":138,"context_line":""},{"line_number":139,"context_line":"Drivers will be implemented inside the oslo.config library, and loaded"},{"line_number":140,"context_line":"through entry points managed by stevedore_ using the namespace"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_52838405","line":137,"in_reply_to":"bff0334d_550bca7b","updated":"2017-04-13 22:48:09.000000000","message":"not just per host, but hostgroups.\n\nsay I have two different compute node pools. one for nodes with gpu\u0027s and one that doesnt. I should be able to label the nodes somewhere to tell it which config it needs to pull from.\n\nNeutron networking needs similar overrides at times.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"7574436e64343db43e6bb15a2cab662542778900","unresolved":false,"context_lines":[{"line_number":166,"context_line":"   This is a significant departure from the current behavior, which"},{"line_number":167,"context_line":"   checks the mutable flags on options and disallows loading new"},{"line_number":168,"context_line":"   settings when the option is not marked as mutable. This needs to be"},{"line_number":169,"context_line":"   discussed."},{"line_number":170,"context_line":""},{"line_number":171,"context_line":"Alternatives"},{"line_number":172,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_6cf1646b","line":169,"updated":"2017-04-10 22:26:22.000000000","message":"After reading up on proposed changes to logging to prepare for the forum, I realized that one of the main use cases for mutable settings is to change the log level. Doing that requires a callback to reconfigure the log handlers, though, and this proposal would remove those callbacks. We\u0027ll have to address that somehow so we don\u0027t lose the feature.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"5d0097314df94034d8ffa2284fa1fb4e82fee06b","unresolved":false,"context_lines":[{"line_number":166,"context_line":"   This is a significant departure from the current behavior, which"},{"line_number":167,"context_line":"   checks the mutable flags on options and disallows loading new"},{"line_number":168,"context_line":"   settings when the option is not marked as mutable. This needs to be"},{"line_number":169,"context_line":"   discussed."},{"line_number":170,"context_line":""},{"line_number":171,"context_line":"Alternatives"},{"line_number":172,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_fe1c56c2","line":169,"in_reply_to":"3f044301_9b7e8469","updated":"2017-05-05 20:56:04.000000000","message":"Today it works by loading all of the config files a second time, then looping over the options it knows and keeping the values that are mutable and warning for the values that changed but shouldn\u0027t have.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2638fe20ab418410528aadf8528ebb0b66004626","unresolved":false,"context_lines":[{"line_number":166,"context_line":"   This is a significant departure from the current behavior, which"},{"line_number":167,"context_line":"   checks the mutable flags on options and disallows loading new"},{"line_number":168,"context_line":"   settings when the option is not marked as mutable. This needs to be"},{"line_number":169,"context_line":"   discussed."},{"line_number":170,"context_line":""},{"line_number":171,"context_line":"Alternatives"},{"line_number":172,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_dca76da7","line":169,"in_reply_to":"7ffa3b31_b273e84d","updated":"2017-04-17 21:45:00.000000000","message":"It might. Unfortunately that\u0027s a bigger change than what we have in mind here.\n\nAnother option is to retain the existing signaling mechanism, and build a callback into the driver to do some sort of resetting.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":6928,"name":"Ben Nemec","email":"openstack@nemebean.com","username":"bnemec"},"change_message_id":"f515603e8005fbc876e3bcce9f417779857d0a11","unresolved":false,"context_lines":[{"line_number":166,"context_line":"   This is a significant departure from the current behavior, which"},{"line_number":167,"context_line":"   checks the mutable flags on options and disallows loading new"},{"line_number":168,"context_line":"   settings when the option is not marked as mutable. This needs to be"},{"line_number":169,"context_line":"   discussed."},{"line_number":170,"context_line":""},{"line_number":171,"context_line":"Alternatives"},{"line_number":172,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_9b7e8469","line":169,"in_reply_to":"7ffa3b31_dca76da7","updated":"2017-05-05 19:43:29.000000000","message":"That sounds reasonable.  It seems like we should be able to selectively invalidate the option cache in the driver so only mutable options can be reloaded, and requiring a callback from the driver doesn\u0027t seem unreasonable.\n\nThat said, take my comments with a big grain of salt on this because I have only a vague familiarity with how the mutable config stuff works.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"e81f60dcadf39526657c90f644d72bf019190454","unresolved":false,"context_lines":[{"line_number":166,"context_line":"   This is a significant departure from the current behavior, which"},{"line_number":167,"context_line":"   checks the mutable flags on options and disallows loading new"},{"line_number":168,"context_line":"   settings when the option is not marked as mutable. This needs to be"},{"line_number":169,"context_line":"   discussed."},{"line_number":170,"context_line":""},{"line_number":171,"context_line":"Alternatives"},{"line_number":172,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_b273e84d","line":169,"in_reply_to":"bff0334d_6cf1646b","updated":"2017-04-13 22:48:09.000000000","message":"A couple of options. clients can watch etcd for changes.\n\nGenerally I\u0027m for immutable config. Then everything\u0027s in a known state.\n\nlog level is one of the few exceptions.                               So maybe log level really belongs in a different mechanism then config in general?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"e81f60dcadf39526657c90f644d72bf019190454","unresolved":false,"context_lines":[{"line_number":212,"context_line":"credentials, for accessing the database and message bus, are"},{"line_number":213,"context_line":"today. One difference in risk level is that the credentials may make"},{"line_number":214,"context_line":"it possible to read all of the other configuration settings from the"},{"line_number":215,"context_line":"central CMDB."},{"line_number":216,"context_line":""},{"line_number":217,"context_line":"Performance Impact"},{"line_number":218,"context_line":"------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_d2c43443","line":215,"updated":"2017-04-13 22:48:09.000000000","message":"I think there\u0027s a big issue here not really covered.\n\nMost of these sorts of tools are boolean, read or not. So all nodes can read all config. With flat files, each service can be configured separately to provide only whats strictly necessary.\n\nThis may be an acceptable risk to some, and not to others. It should be spelled out explicitly.\n\nTo mitigate this risk, better separation of config and secrets could be performed such that secrets are still in files, and config is seperate. Whether or not this proposal makes it through, that should be done anyway.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":7385,"name":"Thomas Herve","email":"therve@redhat.com","username":"therve"},"change_message_id":"ea89bf341493a88755450ba9283188fa77c34ffa","unresolved":false,"context_lines":[{"line_number":212,"context_line":"credentials, for accessing the database and message bus, are"},{"line_number":213,"context_line":"today. One difference in risk level is that the credentials may make"},{"line_number":214,"context_line":"it possible to read all of the other configuration settings from the"},{"line_number":215,"context_line":"central CMDB."},{"line_number":216,"context_line":""},{"line_number":217,"context_line":"Performance Impact"},{"line_number":218,"context_line":"------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_09ee9187","line":215,"in_reply_to":"7ffa3b31_7863b31b","updated":"2017-04-18 20:04:20.000000000","message":"You\u0027re right that etcd ought to provide enough ACLs capabilities to cover this. See https://github.com/coreos/etcd/blob/master/Documentation/op-guide/authentication.md","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"5df920c8a26e8305661b866501863330896b7c1f","unresolved":false,"context_lines":[{"line_number":212,"context_line":"credentials, for accessing the database and message bus, are"},{"line_number":213,"context_line":"today. One difference in risk level is that the credentials may make"},{"line_number":214,"context_line":"it possible to read all of the other configuration settings from the"},{"line_number":215,"context_line":"central CMDB."},{"line_number":216,"context_line":""},{"line_number":217,"context_line":"Performance Impact"},{"line_number":218,"context_line":"------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_7863b31b","line":215,"in_reply_to":"7ffa3b31_d2c43443","updated":"2017-04-18 12:11:06.000000000","message":"We would investigate using etcd authentication, and maybe one role per service? I think this thing has to be done at the deployment tool level. This specs should be able to access to the right k/v with a specific role though, but that\u0027s it.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"9d745347ba14ea2be3da7bd96f2d697e61de1f39","unresolved":false,"context_lines":[{"line_number":213,"context_line":"today. One difference in risk level is that the credentials may make"},{"line_number":214,"context_line":"it possible to read all of the other configuration settings from the"},{"line_number":215,"context_line":"central CMDB."},{"line_number":216,"context_line":""},{"line_number":217,"context_line":"Performance Impact"},{"line_number":218,"context_line":"------------------"},{"line_number":219,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_46a2414d","line":216,"updated":"2017-05-22 18:43:24.000000000","message":"one question was raised at the Summit about how we store passwords. Some folks asked about Castellan and Barbican support but I guess for a first iteration we can keep using the k/v to store secrets as long as k/v has been secured (auth enabled, ACLs, etc). Maybe mention that some work will be done to investigate the usage of Castellan to leverage secrets into another system.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"9d745347ba14ea2be3da7bd96f2d697e61de1f39","unresolved":false,"context_lines":[{"line_number":224,"context_line":"slower than when reading just from an INI file. We can use caching to"},{"line_number":225,"context_line":"mitigate this impact. Drivers are free to implement their own"},{"line_number":226,"context_line":"optimizations internally (such as fetching all of the keys in a"},{"line_number":227,"context_line":"namespace), but the ability to do so is not assumed in the driver API."},{"line_number":228,"context_line":""},{"line_number":229,"context_line":"Configuration Impact"},{"line_number":230,"context_line":"--------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_013793e5","line":227,"updated":"2017-05-22 18:43:24.000000000","message":"indeed, I think the drivers will have some features that other might not have. I\u0027m also interested by handling the case where k/v is not available on the network (it\u0027s a pretty bad situation in production if etcd is deployed correctly, so I\u0027m not sure we want to address it.)","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":7102,"name":"Thomas Bechtold","email":"thomas.bechtold@chainguard.dev","username":"toabctl"},"change_message_id":"e26e1f20796b2d6b689ac9c063215308d8111cd9","unresolved":false,"context_lines":[{"line_number":229,"context_line":"Configuration Impact"},{"line_number":230,"context_line":"--------------------"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"``--config-file`` is changed to ``--config-url``, with the old name"},{"line_number":233,"context_line":"continuing to work as a deprecated alias. All of the following forms"},{"line_number":234,"context_line":"for specifying configuration files are equivalent."},{"line_number":235,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f044301_e7f00900","line":232,"updated":"2017-05-10 13:35:07.000000000","message":"Maybe it would make sense to unify it to --config ? And then this detects if the given parameter is a file (previously --config-file), a dir (previously --config-dir) or a url (new, or we add --config-url for consistency)","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"7ea4f11f83e75145df7e24e263c33e7a43ba27d9","unresolved":false,"context_lines":[{"line_number":229,"context_line":"Configuration Impact"},{"line_number":230,"context_line":"--------------------"},{"line_number":231,"context_line":""},{"line_number":232,"context_line":"``--config-file`` is changed to ``--config-url``, with the old name"},{"line_number":233,"context_line":"continuing to work as a deprecated alias. All of the following forms"},{"line_number":234,"context_line":"for specifying configuration files are equivalent."},{"line_number":235,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_0a5fd9e2","line":232,"in_reply_to":"3f044301_e7f00900","updated":"2017-05-18 16:00:26.000000000","message":"I just replaced \"file\" in the name with \"url\" because we also have \"--config-dir\", which we\u0027ll keep. I don\u0027t feel that strongly about the name, so if others think it should be just --config that\u0027s fine.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"9d745347ba14ea2be3da7bd96f2d697e61de1f39","unresolved":false,"context_lines":[{"line_number":240,"context_line":"  --config-url file:/path/to/file.conf"},{"line_number":241,"context_line":""},{"line_number":242,"context_line":"Deployers using a CMDB will need to load configuration values into"},{"line_number":243,"context_line":"their CMDB using a native tool. The scheme for each CMDB must be"},{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_c6bad108","line":243,"range":{"start_line":243,"start_character":19,"end_line":243,"end_character":30},"updated":"2017-05-22 18:43:24.000000000","message":"not sure you\u0027re talking about this thing but during the Summit we wanted a single tool to push configuration into the CMDB (probably living in oslo.config) that installers / operators could use (it would take a YAML in entry for example, see what Ben is doing: https://review.openstack.org/#/c/451081/). does it make sense?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":1179,"name":"Clay Gerrard","email":"clay.gerrard@gmail.com","username":"clay-gerrard"},"change_message_id":"d7ac8fb1537cae2df504fd444778fd233c7e997d","unresolved":false,"context_lines":[{"line_number":242,"context_line":"Deployers using a CMDB will need to load configuration values into"},{"line_number":243,"context_line":"their CMDB using a native tool. The scheme for each CMDB must be"},{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":"  etcd://user:password@host:port/unique-client-identifier"},{"line_number":248,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_50209c48","line":245,"updated":"2017-04-08 18:38:36.000000000","message":"I don\u0027t understand :\u0027(\n\nI have two big questions:\n\n  1) why is it easier to deploy a url than a config file\n  2) why is it easier for the url to identify a key value store instead of structured configuration document (i.e. https link to download the .ini file)","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"830ccc3e71773ed798b1909e3974b7d97360c28e","unresolved":false,"context_lines":[{"line_number":242,"context_line":"Deployers using a CMDB will need to load configuration values into"},{"line_number":243,"context_line":"their CMDB using a native tool. The scheme for each CMDB must be"},{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":"  etcd://user:password@host:port/unique-client-identifier"},{"line_number":248,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_bf80d3c0","line":245,"in_reply_to":"bff0334d_50209c48","updated":"2017-04-09 19:02:10.000000000","message":"The URL is small enough to be passed into the container via an environment variable. The config file has to either be written to a volume or baked into the image.\n\nThe URL could also download the INI file, and that could certainly be a separate driver. For container deployments, the expectation is that etcd is already present and already being used for configuration.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"7d99193c90f3955f751bc456ebe7a8125fe02ed5","unresolved":false,"context_lines":[{"line_number":242,"context_line":"Deployers using a CMDB will need to load configuration values into"},{"line_number":243,"context_line":"their CMDB using a native tool. The scheme for each CMDB must be"},{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":"  etcd://user:password@host:port/unique-client-identifier"},{"line_number":248,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_5f6b071d","line":245,"in_reply_to":"bff0334d_bf80d3c0","updated":"2017-04-09 19:06:51.000000000","message":"Oh, the other thing using a system like etcd would give us over just downloading files is the change to the behavior of mutable options -- operators could just change the options, and services would noticed the change. No more sending signals to services to reload their config.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"c45140e996ce36c16cf5e5847f74fb658b5a345b","unresolved":false,"context_lines":[{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":"  etcd://user:password@host:port/unique-client-identifier"},{"line_number":248,"context_line":""},{"line_number":249,"context_line":"Within the etcd service, individual configuration options will be"},{"line_number":250,"context_line":"organized based on the project, program, configuration group, and"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_2d8eeccf","line":247,"updated":"2017-04-07 23:00:08.000000000","message":"The python-etcd3 client doesn\u0027t seem to support multiple hosts. Should we worry about that in the URL syntax? https://python-etcd3.readthedocs.io/en/latest/usage.html","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"10e0fdae11181b2ae8f203d3cb859e3e1ef4a77a","unresolved":false,"context_lines":[{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":"  etcd://user:password@host:port/unique-client-identifier"},{"line_number":248,"context_line":""},{"line_number":249,"context_line":"Within the etcd service, individual configuration options will be"},{"line_number":250,"context_line":"organized based on the project, program, configuration group, and"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_f279f7fb","line":247,"in_reply_to":"bff0334d_2d8eeccf","updated":"2017-04-08 16:09:51.000000000","message":"In TripleO and in other deployments, HAproxy is used so we could use one host in frontend, and distribute the traffic in the backend (like we do with Redis in Ceilometer). But I agree in the longterm, we might need to touch the library to support the case where folks don\u0027t have loadbalancers.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"e81f60dcadf39526657c90f644d72bf019190454","unresolved":false,"context_lines":[{"line_number":244,"context_line":"documented in order for them to be able to do this. For example, the"},{"line_number":245,"context_line":"etcd driver will expect a configuration source URL using the form::"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":"  etcd://user:password@host:port/unique-client-identifier"},{"line_number":248,"context_line":""},{"line_number":249,"context_line":"Within the etcd service, individual configuration options will be"},{"line_number":250,"context_line":"organized based on the project, program, configuration group, and"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_12f97c7f","line":247,"in_reply_to":"bff0334d_f279f7fb","updated":"2017-04-13 22:48:09.000000000","message":"I\u0027d much rather the masters be able to spelled out explicitly then to deal with yet another load balancer and trying to make it ha.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"e81f60dcadf39526657c90f644d72bf019190454","unresolved":false,"context_lines":[{"line_number":258,"context_line":"\"myname\" (unique-client-identifier) would be stored at::"},{"line_number":259,"context_line":""},{"line_number":260,"context_line":"  /myname/nova/conductor/oslo_messaging_rabbit/rabbit_host"},{"line_number":261,"context_line":""},{"line_number":262,"context_line":"The \"DEFAULT\" group will not be treated as special in this regard, so"},{"line_number":263,"context_line":"a configuration value \"foo\" that needs to be set in the DEFAULT group"},{"line_number":264,"context_line":"for nova conductor would be stored at::"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_b2ef68af","line":261,"updated":"2017-04-13 22:48:09.000000000","message":"/myname being region?","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2638fe20ab418410528aadf8528ebb0b66004626","unresolved":false,"context_lines":[{"line_number":258,"context_line":"\"myname\" (unique-client-identifier) would be stored at::"},{"line_number":259,"context_line":""},{"line_number":260,"context_line":"  /myname/nova/conductor/oslo_messaging_rabbit/rabbit_host"},{"line_number":261,"context_line":""},{"line_number":262,"context_line":"The \"DEFAULT\" group will not be treated as special in this regard, so"},{"line_number":263,"context_line":"a configuration value \"foo\" that needs to be set in the DEFAULT group"},{"line_number":264,"context_line":"for nova conductor would be stored at::"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7ffa3b31_fcf3499d","line":261,"in_reply_to":"7ffa3b31_b2ef68af","updated":"2017-04-17 21:45:00.000000000","message":"\"myname\" is an arbitrary client identifier. It could also be /my_datacenter/my_zone/my_rack/my_host if someone wanted to organize the config that way. The example isn\u0027t clear, but the point is that we combine the part given to us in the configured URL and add the formula-based portion of the path to get the full location of the option\u0027s value.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"9d745347ba14ea2be3da7bd96f2d697e61de1f39","unresolved":false,"context_lines":[{"line_number":265,"context_line":""},{"line_number":266,"context_line":"  /myname/nova/conductor/DEFAULT/foo"},{"line_number":267,"context_line":""},{"line_number":268,"context_line":"We may want to build a tool to read an INI file and publish it to a"},{"line_number":269,"context_line":"CMDB, but that is not part of this spec and would have to be described"},{"line_number":270,"context_line":"separately before being implemented. Deployment tools such as Tripleo"},{"line_number":271,"context_line":"may provide their mechanism for doing this, or contribute to doing the"},{"line_number":272,"context_line":"work through oslo.config to be shared."}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_e64a7553","line":269,"range":{"start_line":268,"start_character":0,"end_line":269,"end_character":4},"updated":"2017-05-22 18:43:24.000000000","message":"yes, it was one of the needs from operators at Summit.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"c45140e996ce36c16cf5e5847f74fb658b5a345b","unresolved":false,"context_lines":[{"line_number":270,"context_line":"separately before being implemented. Deployment tools such as Tripleo"},{"line_number":271,"context_line":"may provide their mechanism for doing this, or contribute to doing the"},{"line_number":272,"context_line":"work through oslo.config to be shared."},{"line_number":273,"context_line":""},{"line_number":274,"context_line":"Developer Impact"},{"line_number":275,"context_line":"----------------"},{"line_number":276,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"bff0334d_cd3b00e9","line":273,"updated":"2017-04-07 23:00:08.000000000","message":"We need to add some information about the drivers providing their own configuration options and command line switches. We need a function in the same class that is going to provide the open() method mentioned earlier to get those, and then we need to register them. We should also expose those to the configuration generator.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"9d745347ba14ea2be3da7bd96f2d697e61de1f39","unresolved":false,"context_lines":[{"line_number":319,"context_line":"* Change mutate_config_files() to discard all of the existing data and"},{"line_number":320,"context_line":"  reload it, without performing any validation. Fix tests and erase"},{"line_number":321,"context_line":"  dead code left by rewriting that feature."},{"line_number":322,"context_line":"* Write the etcd driver."},{"line_number":323,"context_line":""},{"line_number":324,"context_line":"Incubation"},{"line_number":325,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_e62ef5a3","line":322,"updated":"2017-05-22 18:43:24.000000000","message":"it doesn\u0027t have to be in the spec but some things we also want to provide:\n\n* a python tool to push data from INI into etcd for migrations\n* a python tool to push data into etcd from YAML file (what Ben is doing: https://review.openstack.org/#/c/451081/) if it makes sense","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"bcfa37429fbfa94fb3029f9205069877475716a7","unresolved":false,"context_lines":[{"line_number":319,"context_line":"* Change mutate_config_files() to discard all of the existing data and"},{"line_number":320,"context_line":"  reload it, without performing any validation. Fix tests and erase"},{"line_number":321,"context_line":"  dead code left by rewriting that feature."},{"line_number":322,"context_line":"* Write the etcd driver."},{"line_number":323,"context_line":""},{"line_number":324,"context_line":"Incubation"},{"line_number":325,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_fe9c0b0f","line":322,"in_reply_to":"ff0f0b1f_e62ef5a3","updated":"2017-05-22 19:53:49.000000000","message":"We also need to consider the impact on the design for the validator (https://review.openstack.org/#/c/384559/) since that will need a way to find all of the keys in the remote service, not just the ones known based on the configuration options.","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"},{"author":{"_account_id":3153,"name":"Emilien Macchi","email":"emilien@redhat.com","username":"emilienm"},"change_message_id":"99475db945749fa24cc7c69e8065c23b1d37c5cd","unresolved":false,"context_lines":[{"line_number":329,"context_line":"Adoption"},{"line_number":330,"context_line":"--------"},{"line_number":331,"context_line":""},{"line_number":332,"context_line":"TBD"},{"line_number":333,"context_line":""},{"line_number":334,"context_line":"Library"},{"line_number":335,"context_line":"-------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"ff0f0b1f_abb8c8c7","line":332,"updated":"2017-05-19 18:50:07.000000000","message":"This is some experimental work in Puppet OpenStack modules to configure parameters into etcd instead of configuration files:\nhttps://review.openstack.org/#/c/466292/","commit_id":"28eb593811514dcbcd861bd7f90789982eb3ef6c"}],"specs/queens/oslo-config-drivers.rst":[{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"12a9a718eda2fcad46c1db38d84f161b397995ca","unresolved":false,"context_lines":[{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Centralized configuration management is a highly desired feature as it"},{"line_number":16,"context_line":"allows having a single point of control of all the configuration"},{"line_number":17,"context_line":"parameters for OpenStack components. This is especially try for"},{"line_number":18,"context_line":"containerized deployments, where injecting a file into a container"},{"line_number":19,"context_line":"image is more troublesome than passing smaller amounts of data such as"},{"line_number":20,"context_line":"a URL. However, the existing implementation of oslo.config does not"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_3d959787","line":17,"range":{"start_line":17,"start_character":56,"end_line":17,"end_character":59},"updated":"2017-05-24 22:10:16.000000000","message":"?","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"12a9a718eda2fcad46c1db38d84f161b397995ca","unresolved":false,"context_lines":[{"line_number":38,"context_line":"treated as a ``file://`` URL."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"All connection parameters for the driver, including unique identifiers"},{"line_number":41,"context_line":"for the host or service (if needed), must be part of the URL."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Each occurance of ``--config`` adds a new \"source\" to a search"},{"line_number":44,"context_line":"path. When ConfigOpts looks for an option, it goes through the sources"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_3dda3777","line":41,"updated":"2017-05-24 22:10:16.000000000","message":"I\u0027m not sure thats a good idea... you may need a secret to connect to etcd or whatever driver, and maybe those should be able to pull from config.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2ab0deeb3177d5981bbc659ac117a9d270392433","unresolved":false,"context_lines":[{"line_number":38,"context_line":"treated as a ``file://`` URL."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"All connection parameters for the driver, including unique identifiers"},{"line_number":41,"context_line":"for the host or service (if needed), must be part of the URL."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Each occurance of ``--config`` adds a new \"source\" to a search"},{"line_number":44,"context_line":"path. When ConfigOpts looks for an option, it goes through the sources"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_729d63ba","line":41,"in_reply_to":"ff0f0b1f_3dda3777","updated":"2017-05-25 13:39:38.000000000","message":"I\u0027m trying to keep the first version of this simple. When we have a case where we\u0027re sure we need extra configuration, then we\u0027ll have enough information to design a way to map configuration options coming from these drivers to instances of URLs that cause the drivers to be loaded. In the case of etcd connection details, to start I expect those to be passed to a container image via an environment variable or other similar means, so having a single value makes that easier.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"12a9a718eda2fcad46c1db38d84f161b397995ca","unresolved":false,"context_lines":[{"line_number":41,"context_line":"for the host or service (if needed), must be part of the URL."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Each occurance of ``--config`` adds a new \"source\" to a search"},{"line_number":44,"context_line":"path. When ConfigOpts looks for an option, it goes through the sources"},{"line_number":45,"context_line":"in the reverse order they are provided, to be consistent with the"},{"line_number":46,"context_line":"current behavior. Later sources therefore override earlier sources."},{"line_number":47,"context_line":"The first source that provides a configured value for an option causes"},{"line_number":48,"context_line":"the search to end (sources are not \"merged\" internally)."}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_dd621bae","line":45,"range":{"start_line":44,"start_character":42,"end_line":45,"end_character":38},"updated":"2017-05-24 22:10:16.000000000","message":"This info is a partial implmentation detal. it seems to lack a statement like \"and first key\u003dvalue set wins\".\n\nThe next sentence states the important bit though, that last value assigned to a key wins, with the desired implementation. So, I\u0027d just state the desired state, not the implementation detail at all?","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2ab0deeb3177d5981bbc659ac117a9d270392433","unresolved":false,"context_lines":[{"line_number":41,"context_line":"for the host or service (if needed), must be part of the URL."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Each occurance of ``--config`` adds a new \"source\" to a search"},{"line_number":44,"context_line":"path. When ConfigOpts looks for an option, it goes through the sources"},{"line_number":45,"context_line":"in the reverse order they are provided, to be consistent with the"},{"line_number":46,"context_line":"current behavior. Later sources therefore override earlier sources."},{"line_number":47,"context_line":"The first source that provides a configured value for an option causes"},{"line_number":48,"context_line":"the search to end (sources are not \"merged\" internally)."}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_92b11747","line":45,"range":{"start_line":44,"start_character":42,"end_line":45,"end_character":38},"in_reply_to":"ff0f0b1f_dd621bae","updated":"2017-05-25 13:39:38.000000000","message":"I\u0027m not sure why describing the search algorithm is too much detail? It\u0027s key to understanding the override behavior.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"12a9a718eda2fcad46c1db38d84f161b397995ca","unresolved":false,"context_lines":[{"line_number":159,"context_line":"cached values it holds, as well as any open ConfigurationSource"},{"line_number":160,"context_line":"instances. It will then load its configuration sources again, from"},{"line_number":161,"context_line":"scratch. This avoids the need for a cache-flushing API in the"},{"line_number":162,"context_line":"ConfigurationSource class, keeping the drivers simple."},{"line_number":163,"context_line":""},{"line_number":164,"context_line":"The existing behavior for detecting changes to options not configured"},{"line_number":165,"context_line":"as mutable is retained, as is the existing callback system for"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_a855e3fb","line":162,"updated":"2017-05-24 22:10:16.000000000","message":"in the etcd case, this seems like it would put a lot more load on the etcd server then needed. If the driver knew which keys were mutable, it could set watches on the things and automatically pull in new values with much less overhead.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2ab0deeb3177d5981bbc659ac117a9d270392433","unresolved":false,"context_lines":[{"line_number":159,"context_line":"cached values it holds, as well as any open ConfigurationSource"},{"line_number":160,"context_line":"instances. It will then load its configuration sources again, from"},{"line_number":161,"context_line":"scratch. This avoids the need for a cache-flushing API in the"},{"line_number":162,"context_line":"ConfigurationSource class, keeping the drivers simple."},{"line_number":163,"context_line":""},{"line_number":164,"context_line":"The existing behavior for detecting changes to options not configured"},{"line_number":165,"context_line":"as mutable is retained, as is the existing callback system for"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_72eb2312","line":162,"in_reply_to":"ff0f0b1f_a855e3fb","updated":"2017-05-25 13:39:38.000000000","message":"The mutable option system in oslo.config relies on having callbacks to tell the application the values have changed, to let the app reconfigure itself actively. That\u0027s absolutely needed for logging, for example, and would be needed for database or rabbit connection strings, too. So we can\u0027t just rely on passive detection of a change in the option value.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"12a9a718eda2fcad46c1db38d84f161b397995ca","unresolved":false,"context_lines":[{"line_number":205,"context_line":"Security impact"},{"line_number":206,"context_line":"---------------"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"We assume that all remote access would occur over an encrypted"},{"line_number":209,"context_line":"connection. The details of that connnection are left up to the driver"},{"line_number":210,"context_line":"(the most common use HTTPS)."},{"line_number":211,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_686c6bcd","line":208,"updated":"2017-05-24 22:10:16.000000000","message":"and authenticated?","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2ab0deeb3177d5981bbc659ac117a9d270392433","unresolved":false,"context_lines":[{"line_number":205,"context_line":"Security impact"},{"line_number":206,"context_line":"---------------"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"We assume that all remote access would occur over an encrypted"},{"line_number":209,"context_line":"connection. The details of that connnection are left up to the driver"},{"line_number":210,"context_line":"(the most common use HTTPS)."},{"line_number":211,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_72d44348","line":208,"in_reply_to":"ff0f0b1f_686c6bcd","updated":"2017-05-25 13:39:38.000000000","message":"Sure.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":9237,"name":"Kevin Fox","email":"kevin@efox.cc","username":"kfox1111"},"change_message_id":"12a9a718eda2fcad46c1db38d84f161b397995ca","unresolved":false,"context_lines":[{"line_number":239,"context_line":""},{"line_number":240,"context_line":"  --config-file /path/to/file.conf"},{"line_number":241,"context_line":"  --config /path/to/file.conf"},{"line_number":242,"context_line":"  --config file:/path/to/file.conf"},{"line_number":243,"context_line":""},{"line_number":244,"context_line":"Deployers using a CMDB will need to load configuration values into"},{"line_number":245,"context_line":"their CMDB using a native tool. The scheme for each CMDB must be"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_e89d1bbc","line":242,"updated":"2017-05-24 22:10:16.000000000","message":"--config file:///path/to/file.conf","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"2ab0deeb3177d5981bbc659ac117a9d270392433","unresolved":false,"context_lines":[{"line_number":239,"context_line":""},{"line_number":240,"context_line":"  --config-file /path/to/file.conf"},{"line_number":241,"context_line":"  --config /path/to/file.conf"},{"line_number":242,"context_line":"  --config file:/path/to/file.conf"},{"line_number":243,"context_line":""},{"line_number":244,"context_line":"Deployers using a CMDB will need to load configuration values into"},{"line_number":245,"context_line":"their CMDB using a native tool. The scheme for each CMDB must be"}],"source_content_type":"text/x-rst","patch_set":2,"id":"ff0f0b1f_b2bd3b03","line":242,"in_reply_to":"ff0f0b1f_e89d1bbc","updated":"2017-05-25 13:39:38.000000000","message":"Yes, that is also equivalent.","commit_id":"9625df9c9b61a07b5a9f879a0d03d9e984bf1128"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"489d833a218f5acbb2d78bff87899040b1e595e2","unresolved":false,"context_lines":[{"line_number":45,"context_line":"from a given driver are scanned in the order they are returned from"},{"line_number":46,"context_line":"the driver, allowing the driver to control precedence ordering. (This"},{"line_number":47,"context_line":"behavior allows the ``file`` driver to maintain the current order"},{"line_number":48,"context_line":"handling for ``--config-file`` options on the command line.)"},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"When ``ConfigOpts`` looks for an option, it goes through the drivers"},{"line_number":51,"context_line":"in precedence order and then through the associated sources in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7f515b1d_ff613cd4","line":48,"updated":"2017-10-02 12:09:53.000000000","message":"After looking more closely at the way _ConfigFileOpt works I think we can create a new option type that will register a source so that we can have the command line order control the precedence and not mess around with priorities set in the drivers. I will work on adding more detail about that next week, but for now ignore this part of the spec.","commit_id":"1070b1681e4d40c8cad07284260739be5110679a"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"8fdf202bd99fa0c41fb7e7cbf390d00d66de5415","unresolved":false,"context_lines":[{"line_number":45,"context_line":"from a given driver are scanned in the order they are returned from"},{"line_number":46,"context_line":"the driver, allowing the driver to control precedence ordering. (This"},{"line_number":47,"context_line":"behavior allows the ``file`` driver to maintain the current order"},{"line_number":48,"context_line":"handling for ``--config-file`` options on the command line.)"},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"When ``ConfigOpts`` looks for an option, it goes through the drivers"},{"line_number":51,"context_line":"in precedence order and then through the associated sources in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f4b6375_4d2efffa","line":48,"in_reply_to":"3f4b6375_0db927e8","updated":"2017-10-20 20:39:28.000000000","message":"I apologize for not coming back and updating this.\n\nI meant that each driver would register its own command line option using an special action. It will make more sense if you look at the way the existing --config-file option is implemented in http://git.openstack.org/cgit/openstack/oslo.config/tree/oslo_config/cfg.py#n1624\n\nWhen argparse sees --config-file, it invokes ConfigFileAction which adds the new value to the end of a list of files before parsing it. The list is used later when the app tells ConfigOpts to reload the configuration.\n\nWe want to tie in to the same place using a similar action, but instead of just adding a config file name to a list, we want to add a ConfigurationSource instance.\n\nThe --config-file would append a ConfigFileSource object. An option with a name like --secret-store, implemented by the castellan driver,  might append a CastellanConfigurationSource instance to the same list.\n\nDriver classes wouldn\u0027t need to try to express any sense of precedence, because it would come solely from the command line.\n\nDoes that sound feasible?","commit_id":"1070b1681e4d40c8cad07284260739be5110679a"},{"author":{"_account_id":9914,"name":"Ade Lee","email":"alee@redhat.com","username":"alee"},"change_message_id":"71873a702d99865c82a2e6ee82a03cd191e17f13","unresolved":false,"context_lines":[{"line_number":45,"context_line":"from a given driver are scanned in the order they are returned from"},{"line_number":46,"context_line":"the driver, allowing the driver to control precedence ordering. (This"},{"line_number":47,"context_line":"behavior allows the ``file`` driver to maintain the current order"},{"line_number":48,"context_line":"handling for ``--config-file`` options on the command line.)"},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"When ``ConfigOpts`` looks for an option, it goes through the drivers"},{"line_number":51,"context_line":"in precedence order and then through the associated sources in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f4b6375_0db927e8","line":48,"in_reply_to":"7f515b1d_ff613cd4","updated":"2017-10-20 20:25:48.000000000","message":"If I understand this comment correctly then. an ordered list of ConfigurationSources would be defined on the command line.\n\nEach of these ConfigurationSources would be associated with a particular driver.\n\nSo for example, if I want to store passwords in nova.conf and have them read by a Castellan driver, I might have something like this on the command line:\n\n--config-source CastellanDriver:nova.conf \\\n--config-source FileDriver:nova.conf\n\nWhen initializing the _ConfigOpts then, we would then iterate over the ordered list of ConfigurationSources, and call register_options() only for those drivers associated with ConfigurationSources in the list.\n\nWe would also call get_sources() or maybe something like open_source(ConfigurationSource) on the driver.\n\nWhen getting a config value then, we would iterate through the ordered list of ConfigurationResources, and call get() on each driver, perhaps also passing in the ConfigurationResource.\n\nIs this what you meant?  Or did you still want to use an ordered list of drivers as the sorting mechanism?","commit_id":"1070b1681e4d40c8cad07284260739be5110679a"},{"author":{"_account_id":8866,"name":"Raildo Mascena de Sousa Filho","email":"rmascena@redhat.com","username":"raildo"},"change_message_id":"383726f6c66da9a0d79189a390d67ed9e8d8d35b","unresolved":false,"context_lines":[{"line_number":233,"context_line":"Configuration Impact"},{"line_number":234,"context_line":"--------------------"},{"line_number":235,"context_line":""},{"line_number":236,"context_line":"Deployers using a secret stor will need to load configuration values"},{"line_number":237,"context_line":"into their database using a native tool. The scheme for each backend"},{"line_number":238,"context_line":"service must be documented in order for them to be able to do"},{"line_number":239,"context_line":"this."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7f515b1d_30567a08","line":236,"range":{"start_line":236,"start_character":25,"end_line":236,"end_character":29},"updated":"2017-10-05 19:32:14.000000000","message":"nit: store","commit_id":"1070b1681e4d40c8cad07284260739be5110679a"},{"author":{"_account_id":8866,"name":"Raildo Mascena de Sousa Filho","email":"rmascena@redhat.com","username":"raildo"},"change_message_id":"383726f6c66da9a0d79189a390d67ed9e8d8d35b","unresolved":false,"context_lines":[{"line_number":328,"context_line":""},{"line_number":329,"context_line":"References"},{"line_number":330,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":331,"context_line":""},{"line_number":332,"context_line":"* https://etherpad.openstack.org/p/oslo.config_etcd_backend"},{"line_number":333,"context_line":"* https://etherpad.openstack.org/p/tripleo-etcd-transition"},{"line_number":334,"context_line":"* https://etherpad.openstack.org/p/oslo-config-pluggable-cmdb"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7f515b1d_709ab2c5","line":331,"updated":"2017-10-05 19:32:14.000000000","message":"maybe we should add as reference the last PTG etherpad: https://etherpad.openstack.org/p/oslo-ptg-queens","commit_id":"1070b1681e4d40c8cad07284260739be5110679a"},{"author":{"_account_id":22604,"name":"Samuel Pilla","email":"sp516w@att.com","username":"samuel.pilla"},"change_message_id":"6effd1960d8ceeb42236976e11bc6a2f6ee38160","unresolved":false,"context_lines":[{"line_number":48,"context_line":"also listed on the command line."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"A new option type will be defined for specifying configuration"},{"line_number":51,"context_line":"sources. It will be have like a string option except that it will also"},{"line_number":52,"context_line":"have a custom action that causes it to register some data with the"},{"line_number":53,"context_line":"``ConfigOpts`` instance to indicate what driver should be used to load"},{"line_number":54,"context_line":"the source and the option value that identifies the source. This"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f4b6375_71012206","line":51,"range":{"start_line":51,"start_character":19,"end_line":51,"end_character":20},"updated":"2017-10-25 21:04:33.000000000","message":"nit: extra space","commit_id":"2bdb9732cb682702e70a8e402edf39c96f9af36b"},{"author":{"_account_id":10873,"name":"Juan Antonio Osorio Robles","email":"jaosorior@redhat.com","username":"ejuaoso"},"change_message_id":"d82db4e18e0a093959ca30e415c2e627fc3fe9fa","unresolved":false,"context_lines":[{"line_number":45,"context_line":"example, a URL source might be specified using ``--config-url`` but"},{"line_number":46,"context_line":"client certificates needed may be specified using another command line"},{"line_number":47,"context_line":"option (``--config-url-cert``) or within a configuration file that is"},{"line_number":48,"context_line":"also listed on the command line."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"A new option type will be defined for specifying configuration"},{"line_number":51,"context_line":"sources. It will behave like a string option except that it will also"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_7598591a","line":48,"updated":"2017-11-14 12:44:47.000000000","message":"What about services (such as keystone) that run over httpd? it\u0027s not as straight forward to pass CLI arguments in that case. How will it work in those cases?","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":10873,"name":"Juan Antonio Osorio Robles","email":"jaosorior@redhat.com","username":"ejuaoso"},"change_message_id":"deff50cd9f18dc7be149afba8e63da814da89716","unresolved":false,"context_lines":[{"line_number":45,"context_line":"example, a URL source might be specified using ``--config-url`` but"},{"line_number":46,"context_line":"client certificates needed may be specified using another command line"},{"line_number":47,"context_line":"option (``--config-url-cert``) or within a configuration file that is"},{"line_number":48,"context_line":"also listed on the command line."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"A new option type will be defined for specifying configuration"},{"line_number":51,"context_line":"sources. It will behave like a string option except that it will also"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_7bd6f1c5","line":48,"in_reply_to":"1f485f77_21005401","updated":"2017-11-15 05:18:23.000000000","message":"yeah, this (explicitly pointing out the drivers) seems like the simplest approach (even if it\u0027s more verbose). I\u0027m more inclined towards this.","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"40a3b19030c928cb3ac61e27f8a59f24d75da262","unresolved":false,"context_lines":[{"line_number":45,"context_line":"example, a URL source might be specified using ``--config-url`` but"},{"line_number":46,"context_line":"client certificates needed may be specified using another command line"},{"line_number":47,"context_line":"option (``--config-url-cert``) or within a configuration file that is"},{"line_number":48,"context_line":"also listed on the command line."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"A new option type will be defined for specifying configuration"},{"line_number":51,"context_line":"sources. It will behave like a string option except that it will also"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_d20efe39","line":48,"in_reply_to":"1f485f77_7598591a","updated":"2017-11-14 15:51:12.000000000","message":"Excellent question!\n\nThose processes will need to have a local configuration file in the default location. Any option that can be specified on the command line can also be specified in a configuration file, so the options described here will work as configuration settings. I don\u0027t think they will be loaded automatically though, so we will need to loop over the plugins to ask them for sources.","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"54490c767ae52fa0b4549a8432d329873500a497","unresolved":false,"context_lines":[{"line_number":45,"context_line":"example, a URL source might be specified using ``--config-url`` but"},{"line_number":46,"context_line":"client certificates needed may be specified using another command line"},{"line_number":47,"context_line":"option (``--config-url-cert``) or within a configuration file that is"},{"line_number":48,"context_line":"also listed on the command line."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"A new option type will be defined for specifying configuration"},{"line_number":51,"context_line":"sources. It will behave like a string option except that it will also"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_f5c2db58","line":48,"in_reply_to":"1f485f77_7bd6f1c5","updated":"2017-11-15 18:51:08.000000000","message":"One drawback of doing it that way, without having the option trigger the driver on the command line, is that it will always require a local configuration file that at least describes how to find the actual configuration settings. Maybe we can live with that at first and then figure out how to do something better if it becomes a source of confusion.","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"7409be6e627f7a206305512a91e2a7dd8947f3a8","unresolved":false,"context_lines":[{"line_number":45,"context_line":"example, a URL source might be specified using ``--config-url`` but"},{"line_number":46,"context_line":"client certificates needed may be specified using another command line"},{"line_number":47,"context_line":"option (``--config-url-cert``) or within a configuration file that is"},{"line_number":48,"context_line":"also listed on the command line."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"A new option type will be defined for specifying configuration"},{"line_number":51,"context_line":"sources. It will behave like a string option except that it will also"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_21005401","line":48,"in_reply_to":"1f485f77_d20efe39","updated":"2017-11-15 00:22:03.000000000","message":"Maybe instead of triggering based on special option types we should take the approach cinder uses, and have 1 option that lists the backends and then a separate group for each of those with the options that specify how to connect to it, including the driver. So we might have\n\n  [DEFAULT]\n  config_sources \u003d url, castellan\n\n  [url]\n  driver \u003d url\n  source \u003d https://myserver.com/make-me-a-config-file.cgi\n\n  [castellan]\n  driver \u003d castellan\n  mapping_file \u003d /path/to/mapping.ini\n\nThe config file would need to be discovered automatically, so we need to plumb that in as a special case. The other sources would then be loaded in the order specified by the config_sources option, defined in oslo.config.\n\nBefore I work that into the body of the spec, maybe try to poke some holes in it and see if it wouldn\u0027t work.","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":6159,"name":"Flavio Percoco Premoli","display_name":"flaper87","email":"flavio.percoco@flyrlabs.com","username":"flaper87"},"change_message_id":"fb4a8c95d081632ca615c7e890bd3557944e496a","unresolved":false,"context_lines":[{"line_number":267,"context_line":"Configuration Impact"},{"line_number":268,"context_line":"--------------------"},{"line_number":269,"context_line":""},{"line_number":270,"context_line":"Deployers using a secret stor will need to load configuration values"},{"line_number":271,"context_line":"into their database using a native tool. The scheme for each backend"},{"line_number":272,"context_line":"service must be documented in order for them to be able to do"},{"line_number":273,"context_line":"this."}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_3c7d61b1","line":270,"range":{"start_line":270,"start_character":25,"end_line":270,"end_character":29},"updated":"2017-11-12 03:52:02.000000000","message":"nit: store*","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":6159,"name":"Flavio Percoco Premoli","display_name":"flaper87","email":"flavio.percoco@flyrlabs.com","username":"flaper87"},"change_message_id":"fb4a8c95d081632ca615c7e890bd3557944e496a","unresolved":false,"context_lines":[{"line_number":312,"context_line":"Milestones"},{"line_number":313,"context_line":"----------"},{"line_number":314,"context_line":""},{"line_number":315,"context_line":"queens-2"},{"line_number":316,"context_line":""},{"line_number":317,"context_line":"Work Items"},{"line_number":318,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_1c31850c","line":315,"range":{"start_line":315,"start_character":0,"end_line":315,"end_character":8},"updated":"2017-11-12 03:52:02.000000000","message":"Queens 2 is in aboout 2 or 3 weeks. Are we confident this will be implemented by Q2? Would it be better to just target Rocky for this feature?\n\nMerging the spec would already be a good achievement, in my opinion.","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"f76f8c883ce0d24a056df99f053feb2d86cd7891","unresolved":false,"context_lines":[{"line_number":312,"context_line":"Milestones"},{"line_number":313,"context_line":"----------"},{"line_number":314,"context_line":""},{"line_number":315,"context_line":"queens-2"},{"line_number":316,"context_line":""},{"line_number":317,"context_line":"Work Items"},{"line_number":318,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":6,"id":"1f485f77_a9d0f0ca","line":315,"range":{"start_line":315,"start_character":0,"end_line":315,"end_character":8},"in_reply_to":"1f485f77_1c31850c","updated":"2017-11-12 16:40:47.000000000","message":"At this point aiming for rocky is probably safer. We should be able to do some work this cycle, and land full support next cycle. Let\u0027s wait to see what the AT\u0026T folks say before updating the target here.","commit_id":"da6da0ac2d3d21407d0d6573b365df7f75236d86"}]}
