)]}'
{"specs/version0.5/amphora-driver-interface.rst":[{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"b8778bc0540d917bdf504b89b7be5cbffcf0c138","unresolved":false,"context_lines":[{"line_number":24,"context_line":""},{"line_number":25,"context_line":".. code:: python"},{"line_number":26,"context_line":""},{"line_number":27,"context_line":"    class AmphoraeDriver(object):"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_8d073218","line":27,"updated":"2014-09-16 18:52:54.000000000","message":"At the risk of causing another naming debate, I\u0027m going to mention it anyway.  This name makes it sound like its controlling the Amphorae, and not the actual load balancing software (haproxy).  I know I named the blueprint amphora-driver-interface, but if we name this appropriately, then the method names will make sense within its context.  There is another driver interface blueprint called the compute-driver-interface, in which that will be responsible for spinning up/down amphorae.  Maybe SoftwareLoadBalancerDriver would be a better name (albeit a bit long).\n\nAlso, are you not using abstract base class because you want the OPTIONAL methods to throw the NotImplementedError only when they\u0027re called?","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":24,"context_line":""},{"line_number":25,"context_line":".. code:: python"},{"line_number":26,"context_line":""},{"line_number":27,"context_line":"    class AmphoraeDriver(object):"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_6202f470","line":27,"in_reply_to":"fa98f980_44f423cf","updated":"2014-09-17 20:03:28.000000000","message":"An amphorae can contain multiple load balancers (as we decided with the M:N relationship and this driver is meant to manage all the LBs on an amphora so I would rather stick with that name and name the the other one AmphoraCreationDriver...","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"c708bc56944ddabc25363af59b38d67bd38675c8","unresolved":false,"context_lines":[{"line_number":24,"context_line":""},{"line_number":25,"context_line":".. code:: python"},{"line_number":26,"context_line":""},{"line_number":27,"context_line":"    class AmphoraeDriver(object):"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_942f2888","line":27,"in_reply_to":"fa98f980_8d073218","updated":"2014-09-17 09:20:06.000000000","message":"I don\u0027t think this abstraction is meant to be haproxy-specific. It\u0027s the driver that will be haproxy-specific, and what German\u0027s defining here is the interface between the controller and the driver.  So calling this the AmphoraeDriver class is probably actually accurate.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"9bffffe0d6e276be8665ffaed4e3c2b8b2081594","unresolved":false,"context_lines":[{"line_number":24,"context_line":""},{"line_number":25,"context_line":".. code:: python"},{"line_number":26,"context_line":""},{"line_number":27,"context_line":"    class AmphoraeDriver(object):"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_44f423cf","line":27,"in_reply_to":"fa98f980_942f2888","updated":"2014-09-17 16:29:58.000000000","message":"I\u0027m not saying it needs to be haproxy-specific, just that it needs to be load balancing software centric.  If the interfaec class is SoftwareLoadBalancerDriver, then its obvious that implementations of it would be HaproxyDriver, NginxDriver, etc.  AmphoraeDriver to me sounds like its responsible for the actual creation of the amphorae.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":26,"context_line":""},{"line_number":27,"context_line":"    class AmphoraeDriver(object):"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"        def update(self, lbs):"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_7e100f59","line":29,"updated":"2014-09-16 00:46:17.000000000","message":"Could you elaborate on what you mean by \u0027custom logger\u0027? Are you talking about a custom log line for the haproxy.cfg file or something else?","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":26,"context_line":""},{"line_number":27,"context_line":"    class AmphoraeDriver(object):"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"        def update(self, lbs):"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_021128af","line":29,"in_reply_to":"fa98f980_7e100f59","updated":"2014-09-17 20:03:28.000000000","message":"Ideally the driver will write log output to getLogger().error(\"Bla Bla\") This is a way to inject a logger instead of using a LOGGER constant at the beginning of each driver...","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"        def update(self, lbs):"},{"line_number":33,"context_line":"            #update the amphorae with a new configuration for each load balancer specified in the array lbs"},{"line_number":34,"context_line":"            raise NotImplementedError"},{"line_number":35,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_feef9fae","line":32,"updated":"2014-09-16 00:46:17.000000000","message":"So, these CRUD-like methods that work against the load balancer object should probably be named such that it\u0027s apparent from the method name they\u0027re doing operations on the LB instead of the Amphora as a whole. Same goes for the CRUD-like operations for a listener (as requested below).","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"c708bc56944ddabc25363af59b38d67bd38675c8","unresolved":false,"context_lines":[{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"        def update(self, lbs):"},{"line_number":33,"context_line":"            #update the amphorae with a new configuration for each load balancer specified in the array lbs"},{"line_number":34,"context_line":"            raise NotImplementedError"},{"line_number":35,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_145b382f","line":32,"in_reply_to":"fa98f980_a3f16203","updated":"2014-09-17 09:20:06.000000000","message":"One or the other needs to happen, yes.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"b8778bc0540d917bdf504b89b7be5cbffcf0c138","unresolved":false,"context_lines":[{"line_number":29,"context_line":"        def getLogger(self):"},{"line_number":30,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"        def update(self, lbs):"},{"line_number":33,"context_line":"            #update the amphorae with a new configuration for each load balancer specified in the array lbs"},{"line_number":34,"context_line":"            raise NotImplementedError"},{"line_number":35,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_a3f16203","line":32,"in_reply_to":"fa98f980_feef9fae","updated":"2014-09-16 18:52:54.000000000","message":"@Stephen: If the class name is changed to something more specific (but still generic) then these methods should be able to stay the same name since the context would be within this method.  As for separate methods for listeners, I don\u0027t agree for the reasons stated below.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":45,"context_line":"            #delete a load balancer on the amphora"},{"line_number":46,"context_line":"            raise NotImplementedError"},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"        def delete(self, amphora):"},{"line_number":49,"context_line":"            #delete all load balancers and the amphora"},{"line_number":50,"context_line":"            raise NotImplementedError"},{"line_number":51,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_1e6543ae","line":48,"updated":"2014-09-16 00:46:17.000000000","message":"Need a different name for this or the previous method. (Can\u0027t repeat method names, yo.)","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":54,"context_line":"            #some information might come from querying the amphora"},{"line_number":55,"context_line":"            raise NotImplementedError"},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"        def archive_logs(self, loadbalancer, listener\u003dALL, target\u003dSWIFT):"},{"line_number":58,"context_line":"            #archives logs for a specific load balancer, listener, etc."},{"line_number":59,"context_line":"            raise NotImplementedError"},{"line_number":60,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_be21974e","line":57,"updated":"2014-09-16 00:46:17.000000000","message":"Seeing as how we have not really gotten any concrete requirements around logging from anybody in this discussion yet, I would rather we leave log-related commands out of the API interface until we have said requirements. (Otherwise we may end up implementing something that nobody actually uses.)","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":54,"context_line":"            #some information might come from querying the amphora"},{"line_number":55,"context_line":"            raise NotImplementedError"},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"        def archive_logs(self, loadbalancer, listener\u003dALL, target\u003dSWIFT):"},{"line_number":58,"context_line":"            #archives logs for a specific load balancer, listener, etc."},{"line_number":59,"context_line":"            raise NotImplementedError"},{"line_number":60,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_025f6876","line":57,"in_reply_to":"fa98f980_be21974e","updated":"2014-09-17 20:03:28.000000000","message":"Somehow making logs accessible to end users is an important requirement so I would like to leave that in there at the risk of having to change it later - so it doesn\u0027t get forgotten. But we all have different ways to remind us... so I could be persuaded to scrap it.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":59,"context_line":"            raise NotImplementedError"},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"        def get_metrics(self, amphora):"},{"line_number":62,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them straight to ceilometer"},{"line_number":63,"context_line":"            raise NotImplementedError"},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"        def get_statistics(self, amphora):"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_3e912747","line":62,"updated":"2014-09-16 00:46:17.000000000","message":"We should probably decide what behavior we allow here and enforce that from all drivers. Should the controller / driver be responsible for talking to ceilometer or the amphora?  (Can we even guarantee the amphora will always be able to talk to ceilometer? Has anyone here investigated what that interface looks like?)","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":59,"context_line":"            raise NotImplementedError"},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"        def get_metrics(self, amphora):"},{"line_number":62,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them straight to ceilometer"},{"line_number":63,"context_line":"            raise NotImplementedError"},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"        def get_statistics(self, amphora):"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_22cb8c1f","line":62,"in_reply_to":"fa98f980_3e912747","updated":"2014-09-17 20:03:28.000000000","message":"I wanted to leave the door open for https://blueprints.launchpad.net/ceilometer/+spec/paas-event-format-for-ceilometer -- I think the interface is flexible enough to handle if we get a NotImplemented to assume the metrics will come from somewhere else.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":62,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them straight to ceilometer"},{"line_number":63,"context_line":"            raise NotImplementedError"},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"        def get_statistics(self, amphora):"},{"line_number":66,"context_line":"            #return statistics, e.g, listener health, node health, etc."},{"line_number":67,"context_line":"            raise NotImplementedError"},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_5e8d0b30","line":65,"updated":"2014-09-16 00:46:17.000000000","message":"So \"statistics\" as a name sounds a lot more like the metrics described in the previous method. Perhaps we could call this \"get_status\" instead?\n\nIs the intent here to return the listener health, node health, etc. for *all* load balancers on the amphora? That can be relatively expensive to gather on the fly, so how often do we think this method is going to be called?\n\nPerhaps we want to have a hierarchy for status here:  For example, asking for the status of the amphora gives the overall status of the amphora plus the overall status of the load balancers on it. Asking for the status of a specific load balancer gives the overall status of the load balancer and overall status of the listeners, etc.\n\nAlso keep in mind there\u0027s a difference between operational status and provisional status, as discussed on the mailing list.\n\nIn any case, this method description seems a bit too generic unless we\u0027re planning on dumping essentially the status of everything with one method call.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":62,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them straight to ceilometer"},{"line_number":63,"context_line":"            raise NotImplementedError"},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"        def get_statistics(self, amphora):"},{"line_number":66,"context_line":"            #return statistics, e.g, listener health, node health, etc."},{"line_number":67,"context_line":"            raise NotImplementedError"},{"line_number":68,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_8200b836","line":65,"in_reply_to":"fa98f980_5e8d0b30","updated":"2014-09-17 20:03:28.000000000","message":"I don\u0027t really see a reason why we wouldn\u0027t dump everything at regular intervals. I think we don\u0027t have a requirement to report status for LB1 more frequent than LB2. But we can move to the json format discussed below to make it more palatable.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":66,"context_line":"            #return statistics, e.g, listener health, node health, etc."},{"line_number":67,"context_line":"            raise NotImplementedError"},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"        def get_diagnostics(self, amphora):"},{"line_number":70,"context_line":"            #run some self tests, e.g. connectivity, etc. - to determine if that amphorae is healthy"},{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_9e9a1360","line":69,"updated":"2014-09-16 00:46:17.000000000","message":"How does this differ significantly from the \u0027info\u0027 or \u0027get_statistics\u0027 methods above? Specifically, \u0027info\u0027 is probably already going to need to do some information gathering before it can report said info.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":66,"context_line":"            #return statistics, e.g, listener health, node health, etc."},{"line_number":67,"context_line":"            raise NotImplementedError"},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"        def get_diagnostics(self, amphora):"},{"line_number":70,"context_line":"            #run some self tests, e.g. connectivity, etc. - to determine if that amphorae is healthy"},{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_22c0ecdf","line":69,"in_reply_to":"fa98f980_9e9a1360","updated":"2014-09-17 20:03:28.000000000","message":"Well, this is triggering diagnostics on the amphora which might be more expensive then looking in the process table. In libra for instance we try to connect to well defined ips to make sure we have network connectivity to them.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The referenced lb is a load balancer object as described in our model. The model is detached from the DB so the driver can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":74,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at this juncture. This might be added in later versions."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"Exception Model"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_bed277fd","line":74,"updated":"2014-09-16 00:46:17.000000000","message":"Ok, if this is the case, then we need the CRUD methods above to work against the \u0027Listener\u0027 since there will be one haproxy process / one haproxy config per listener.\n\nIs there a particular reason we\u0027re against the driver having DB access? It seems to me that if we\u0027re tracking state information (eg. Member up/down) in the database at all-- which we will probably need to do-- it\u0027s going to be useful to have the driver able to update the database...  unless we\u0027re specifically going to add call-backs (mixins) for the driver to communicate these kinds of spontaneous status changes to the controller in some way.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"9bffffe0d6e276be8665ffaed4e3c2b8b2081594","unresolved":false,"context_lines":[{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The referenced lb is a load balancer object as described in our model. The model is detached from the DB so the driver can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":74,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at this juncture. This might be added in later versions."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"Exception Model"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_3f3c8cbf","line":74,"in_reply_to":"fa98f980_140df837","updated":"2014-09-17 16:29:58.000000000","message":"Yeah that is something I forgot about.  I would still rather it not be endorsed by the driver, but that might end up being the best solution, and if someone creates and driver that doesn\u0027t do an instance per listener, then they just won\u0027t implement the listener methods.  The other solutions are 2) as you suggested the amphora tests that the config has changed (eww) 3) we add attributes to the listeners and load balancer models to say when they\u0027ve been changed (less eww, so ew).\n\nI think no DB access is worth doing.  I\u0027m sure this is something we can re-evaluate if it gets out of control and we\u0027re having to do crazy things.  I\u0027m sure others have opinions on it.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The referenced lb is a load balancer object as described in our model. The model is detached from the DB so the driver can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":74,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at this juncture. This might be added in later versions."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"Exception Model"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_a535c6dc","line":74,"in_reply_to":"fa98f980_3f3c8cbf","updated":"2014-09-17 20:03:28.000000000","message":"I agree with Brandon - no DB access to our tables. I am a little bit lenient if a driver decides to have their own shema/db vut that\u0027s a future discussion.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"c708bc56944ddabc25363af59b38d67bd38675c8","unresolved":false,"context_lines":[{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The referenced lb is a load balancer object as described in our model. The model is detached from the DB so the driver can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":74,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at this juncture. This might be added in later versions."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"Exception Model"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_140df837","line":74,"in_reply_to":"fa98f980_4346f650","updated":"2014-09-17 09:20:06.000000000","message":"So, if we\u0027re pushing an entire load balancer configuration to the driver with even the smallest sub-entity update, how is the driver to know when it doesn\u0027t actually need to go out an change a configuration on the Amphora?  (For example, if I update a member of pool A that\u0027s part of listener A, the driver shouldn\u0027t need to manipulate the haproxy.cfg for listener B, which happens to be on the same load balancer as listener A. Yet if we\u0027re pushing the whole load balancer config each time... ?)\n\nI suppose it\u0027s possible to generate new haproxy.cfg\u0027s for each of the listeners, and push them out, and leave it up to the Amphora whether it decides to restart the haproxy services for a given listener (ie. don\u0027t, if the config just pushed is identical to the one already in place). Are you advocating this?\n\nAlso, suppose the user deletes listener A in the above example (but not the load balancer or listener B). How does the driver handle this?\n\nI\u0027m not yet convinced it\u0027s unreasonable to expect drivers to treat each listener as a separate entity-- after all this is the behavior that we\u0027re going to have with the haproxy driver (this is the implication from the \"one haproxy instance per listener\" debate we had and voted on).\n\nAs far as not allowing the drivers to have DB access: In order to enforce this you will need to send anything the driver would possibly need to know from the DB with even the smallest updates, because without DB access the driver is effectively stateless. Be aware that means the controller\u003c-\u003edriver interface is likely to see a lot of changes as different types of drivers are developed.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"b8778bc0540d917bdf504b89b7be5cbffcf0c138","unresolved":false,"context_lines":[{"line_number":71,"context_line":"            raise NotImplementedError"},{"line_number":72,"context_line":""},{"line_number":73,"context_line":"The referenced lb is a load balancer object as described in our model. The model is detached from the DB so the driver can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":74,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at this juncture. This might be added in later versions."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"Exception Model"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_4346f650","line":74,"in_reply_to":"fa98f980_bed277fd","updated":"2014-09-16 18:52:54.000000000","message":"@Stephen: If this interface has separate listener methods then that\u0027s kind of enforcing that all drivers do separate listeners per load balancer instance.  So I think another solution is needed.\n\nRegarding no DB access.  I agree with German on this.  I don\u0027t like the fact that in neutron lbaas that the drivers are responsible for some of the DB calls.  It leads to inconsistent DB access across many drivers.  To me, the only thing this driver needs to worry about is the manipulating the software load balancers on amphorae. I would prefer that the only components to have access to the database are the API and controller.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":77,"context_line":"Exception Model"},{"line_number":78,"context_line":"---------------"},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"The driver is expected to raise the follow well defined exceptions"},{"line_number":81,"context_line":""},{"line_number":82,"context_line":"* NotImplementedError - this functionality is not implemented/not supported"},{"line_number":83,"context_line":"* AmphoraDriverError - a super class for all other exceptions and the catch all if no specific exception can be determined."}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_7e33af61","line":80,"updated":"2014-09-16 00:46:17.000000000","message":"Nit: following","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":85,"context_line":"    * DeleteFailed - the amphora couldn\u0027t be deleted"},{"line_number":86,"context_line":"    * AmphoraLoadBalancerError - errors to a specific load balancer"},{"line_number":87,"context_line":"        * DeleteFailed - this load balancer couldn\u0027t be deleted"},{"line_number":88,"context_line":"        * SuspendFaied - this load balancer couldn\u0027t be suspended"},{"line_number":89,"context_line":"        * EnableFailed - this load balancer couldn\u0027t be enabled"},{"line_number":90,"context_line":"        * ArchiveException - couldn\u0027t argcive the logs"},{"line_number":91,"context_line":"            * TargetException - the target is not accessible"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_9e0db391","line":88,"updated":"2014-09-16 00:46:17.000000000","message":"Nit: Failed","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":96,"context_line":"    * UnauthorizedException - the driver can\u0027t access the amphora"},{"line_number":97,"context_line":"    * StatisticsException - gathering statistics failed"},{"line_number":98,"context_line":"    * TimeOutException - contacting the amphora timed out"},{"line_number":99,"context_line":"    * UnavailableException - the amphora is temporary unavailable"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Health and Stat Mixin"},{"line_number":102,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_fe4b1fae","line":99,"updated":"2014-09-16 00:46:17.000000000","message":"Will we be communicating operational failures (eg. Amphora  / listener / member, etc. are down) via exceptions?","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":96,"context_line":"    * UnauthorizedException - the driver can\u0027t access the amphora"},{"line_number":97,"context_line":"    * StatisticsException - gathering statistics failed"},{"line_number":98,"context_line":"    * TimeOutException - contacting the amphora timed out"},{"line_number":99,"context_line":"    * UnavailableException - the amphora is temporary unavailable"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Health and Stat Mixin"},{"line_number":102,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_0581324d","line":99,"in_reply_to":"fa98f980_fe4b1fae","updated":"2014-09-17 20:03:28.000000000","message":"At least provisioning errors. When an LB goes down during operations that might be part of health statistics (unless the agent on the amphora is unresponsive as well).","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":111,"context_line":"        def update_health(health):"},{"line_number":112,"context_line":"            #map: {\"amphora-status\":HEALTHY, loadbalancers: {\"loadbalancer-id\": {\"loadbalancer-status\": HEALTHY, \"listeners\":{\"listener-id\":{\"listener-status\":HEALTHY, \"nodes\":{\"node-id\":HEALTHY, ...}}, ...}, ...}}"},{"line_number":113,"context_line":"            #awesome update code"},{"line_number":114,"context_line":"            pass"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"    class StatsMixIn(object):"},{"line_number":117,"context_line":"        def update_stats(stats):"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_fef95f4e","line":114,"updated":"2014-09-16 00:46:17.000000000","message":"One packet for the health of everything? Or are you thinking that any edge changes simply have their data in the JSON, and anything not listed is assumed to be \u0027not changed\u0027?  (Which probably is a good way to do this.)","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":111,"context_line":"        def update_health(health):"},{"line_number":112,"context_line":"            #map: {\"amphora-status\":HEALTHY, loadbalancers: {\"loadbalancer-id\": {\"loadbalancer-status\": HEALTHY, \"listeners\":{\"listener-id\":{\"listener-status\":HEALTHY, \"nodes\":{\"node-id\":HEALTHY, ...}}, ...}, ...}}"},{"line_number":113,"context_line":"            #awesome update code"},{"line_number":114,"context_line":"            pass"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"    class StatsMixIn(object):"},{"line_number":117,"context_line":"        def update_stats(stats):"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_25863666","line":114,"in_reply_to":"fa98f980_fef95f4e","updated":"2014-09-17 20:03:28.000000000","message":"I was thinking the later.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":117,"context_line":"        def update_stats(stats):"},{"line_number":118,"context_line":"            #uses map {\"loadbalancer-id\":{\"listener-id\": {\"bytes-in\": 123, \"bytes_out\":123, \"active_connections\":123, \"total_connections\", 123}, ...}"},{"line_number":119,"context_line":"            #awesome update code and code to send to ceilometer"},{"line_number":120,"context_line":"            pass"},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"Things a good driver should do:"},{"line_number":123,"context_line":"-------------------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_3e65a71d","line":120,"updated":"2014-09-16 00:46:17.000000000","message":"Does it make sense to list the exact things we\u0027re going to be gathering statistics on, as well as the data type being passed?  eg. \"bytes_in\" is probably a counter, whereas \"current_connections\" is probably a gauge.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3041ae441964073cb3ba2b3b0e5dbedff7d7ec4a","unresolved":false,"context_lines":[{"line_number":117,"context_line":"        def update_stats(stats):"},{"line_number":118,"context_line":"            #uses map {\"loadbalancer-id\":{\"listener-id\": {\"bytes-in\": 123, \"bytes_out\":123, \"active_connections\":123, \"total_connections\", 123}, ...}"},{"line_number":119,"context_line":"            #awesome update code and code to send to ceilometer"},{"line_number":120,"context_line":"            pass"},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"Things a good driver should do:"},{"line_number":123,"context_line":"-------------------------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_e58b4e66","line":120,"in_reply_to":"fa98f980_3e65a71d","updated":"2014-09-17 20:03:28.000000000","message":"I want to keep it extensible for future versions.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d8ca8e8a4454300206a057e4f0704df7aa45029","unresolved":false,"context_lines":[{"line_number":123,"context_line":"-------------------------------"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":" * Non blocking IO - throw an appropriate exception instead to wait forever; use timeouts on sockets"},{"line_number":126,"context_line":" * We might employ a circuit breaker to insulate driver problems from controller problems [1]"},{"line_number":127,"context_line":" * Use appropriate logging"},{"line_number":128,"context_line":" * Use the preferred threading model"},{"line_number":129,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_7eb48f81","line":126,"updated":"2014-09-16 00:46:17.000000000","message":"Not a bad idea--  but if the driver code is that bad, why would anyone run it?","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"69f40735b32d304a9ebb9cec29180b49dc8e22bf","unresolved":false,"context_lines":[{"line_number":123,"context_line":"-------------------------------"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":" * Non blocking IO - throw an appropriate exception instead to wait forever; use timeouts on sockets"},{"line_number":126,"context_line":" * We might employ a circuit breaker to insulate driver problems from controller problems [1]"},{"line_number":127,"context_line":" * Use appropriate logging"},{"line_number":128,"context_line":" * Use the preferred threading model"},{"line_number":129,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"fa98f980_1cafc87e","line":126,"in_reply_to":"fa98f980_7eb48f81","updated":"2014-09-18 05:15:04.000000000","message":"Could be some weird edge/corner case undetected until it has been sitting in a production environment for a while and then decides to rear its ugly head.","commit_id":"a4748c8d443f57d88ebd291a4ed4e2001a80b776"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"2db2804c45f37f084ed4e00499f10f302b6d5a4e","unresolved":false,"context_lines":[{"line_number":9,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":10,"context_line":"https://blueprints.launchpad.net/octavia/+spec/amphora-driver-interface"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"This blueprint describes how a driver will interface with the controller."},{"line_number":13,"context_line":"It will describe the base class and other"},{"line_number":14,"context_line":"classes required. It will not describe the REST interface needed to"},{"line_number":15,"context_line":"talk to an amphora nor"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_98ad649e","line":12,"updated":"2014-09-19 03:34:42.000000000","message":"While this kind of line spacing does meet pep8, it does look pretty ugly in raw text. Part of the point of the .rst format is you can have good looking documentation both in the raw text, and if a form that\u0027s easily parseable into HTML, PDF, or other documentation. German-- can you adjust your text editor settings to simply do 79-character word wrapping so we don\u0027t have lines that are unnecessarily short like this?","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"24eb1a06a05261de5e369bfed1a76b54bdbb35b9","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_bcd6bcef","line":38,"updated":"2014-09-18 05:14:19.000000000","message":"@German: you didn\u0027t add listener methods.  Do you have an opinion/solution to the issues discussed in the previous PS?","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d97047174626cb37b5d844888026dfc618c4a8a","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_cfef47af","line":38,"in_reply_to":"da9df570_c052bf12","updated":"2014-09-23 03:44:16.000000000","message":"I really dislike the idea of an implicit delete of a listener like that. It seems to me to be too prone to error, especially if pushing out haproxy configurations to the amphora is done one configuration at a time in separate back-end API commands (as it probably ought to be to keep these commands and any errors they generate simple and relatively \"dumb\").\n\nI also generally dislike the idea of relying on state on the amphora for what the driver has to do. :/\n\nI don\u0027t understand why you would think we would need a single method for pushing out a complete loadbalancer configuration when failing over: What\u0027s wrong with iterating through each of the listeners on the loadbalancer?\n\nWhy do you think that pushing out individual listener configurations is more complex than pushing out a whole load balancer configuration to the driver (and having the driver or amphora detect when a listener has changed or has been deleted)? It means the controller has to be slightly more intelligent, but the driver does not. Nor does the amphora.\n\nIf a given vendor implementation isn\u0027t able to deal with just the data corresponding to a listener (ie. if they have to push out an entire loadbalancer config at once)... well, first of all this is not congruent to what we\u0027re going to be doing with the reference haproxy amphora, and therefore isn\u0027t just a drop-in replacement for the haproxy amphora. And anyway, if they have to do this, then THEY could potentially tackle the problem of storing state on their amphora, and generate the whole loadbalancer configuration given the single listener that has changed... or they can change their implementation to actually mimic the haproxy amphora behavior here (as they probably should). :P Again, in our constitution we give priority to the open source solution on decisions like this.\n\nGerman: I get the distinct impression that you\u0027re planning on just writing a driver that pushes the whole loadbalancer configuration out to the Amphora (despite our public discussion, voting, and decision on this) so you have an easier time writing a driver that works with the virtual load balancer solution you\u0027ve already got in place (Libra).\n\nIf this is what you\u0027re doing, please note that this doesn\u0027t make sense, since Octavia is a whole command-and-control system for virtualized load balancers that is very likely going to be incompatible with how you\u0027ve got things working in Libra. That\u0027s not to say we don\u0027t want to learn from your experience with Libra-- I\u0027m hoping by working together we can come up with a system that works better than any of us have created in the past (at Blue Box, we\u0027re already making compromises which make Octavia both incompatible with and probably improved upon our BLBv2 load balancer product). Even if you push an agenda like this, it\u0027s very, very unlikely that the group here working on Octavia is going to create something that works exactly similar to Libra in the end, and therefore there will be incompatibilities. I think it would be better if we learned from Libra, but didn\u0027t try to jam Octavia into a Libra-shaped mold.\n\nIn any case: I am very uncomfortable with the implicit listener delete. I would like to see an explicit listener delete method in any case. (As well as methods for starting / stopping an individual listener so we can over-ride a lazy restart.)\n\nTo be frank, I\u0027m not sure we need \"loadbalancer\" methods at all since this is really just a logical construct that links VIPs to Listeners to Amphorae. Rather, I think we would both do well with VIP-related methods (ie. create and delete-- telling the amphora to bind / unbind from a given VIP) and listener-related methods. This would also keep the domain of influence separate. (ie. \"loadbalancer\" methods overlap with \"listener\" methods if both are included, inasmuch as loadbalancer methods manipulate listeners.)\n\nThe one thing I do concede here is that pushing out the whole loadbalancer configuration, and relying on an implicit delete is potentially more idempotent than an explicit listener delete for a few specific kinds of predictable failures. (ie. a missed explicit delete command potentially leaves an unwanted listener lying around unless we build a verification step in there somewhere to make sure we don\u0027t have orphaned listeners. Though a missed explicit delete would presumably be caught with an error we can actually do something about.) But pushing a whole loadbalancer config to the driver has more difficult error scenarios to deal with simply because pushing a whole load balancer configuration to an amphora does a lot more than just pushing or deleting an haproxy config. (Resulting in scenarios where the push of one haproxy config results in all the listeners for a given loadbalancer ending up in the \u0027ERROR\u0027 state because we can\u0027t report a more specific error... and depending on what the amphora does can still leave orphaned listeners lying around in many error scenarios.)","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"5eb8c8b3d505c793f747f5bb6f121d5a44862820","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_6a7c8921","line":38,"in_reply_to":"da9df570_cfef47af","updated":"2014-09-23 03:55:50.000000000","message":"One note about the above: I do argue vehemently on these matters, of course-- I want Octavia to be the best thing it can be and I specifically try to kill what I think are bad ideas before they gain much traction. But I could totally be wrong. What I mean by the above is:\n\nConvince me why your idea is better, and I will argue just as vehemently for your idea as I argued against it. I will absolutely switch sides if you can back up your ideas and/or shoot down mine.\n\nPlease also note that I will tend to argue very strongly against a point if I smell any kind of agenda there other than \"make Octavia the best thing it can be for everyone.\"  :) (So, part of your justification above might be to show me why this isn\u0027t an agenda to shoehorn Libra into the mix somehow. ;) )","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"1e05eba58fe706e48887e2b857e9ec55ccb3282f","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_5e7bef7a","line":38,"in_reply_to":"fa98f980_56c2f953","updated":"2014-09-18 19:48:33.000000000","message":"So are you suggesting the code running on the amphora will just update all the haproxy instances even if only one listener has been updated?  Or do a config compare to see which one\u0027s changed?","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"24662756c1ba6bf247fc0b93ee4723e044fe873e","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_bed4dbbb","line":38,"in_reply_to":"fa98f980_5e7bef7a","updated":"2014-09-18 19:54:43.000000000","message":"In my experience people mostly have no more than two listeners so updating both or comparing isn\u0027t that much of a deal. But I will leave that as an implementation detail on that other blueprint ;-)","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"73c0a71ba319aa2e7e766b90d1e9939004681ec9","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_c052bf12","line":38,"in_reply_to":"fa98f980_98e9e4b9","updated":"2014-09-22 21:59:27.000000000","message":"The other blueprint would of course be the amphora driver https://blueprints.launchpad.net/octavia/+spec/haproxy-amphora-driver\n\n\" A single haproxy configuration corresponds with a Listener, not a Loadbalancer.\" It looks to me we are leaking details of the underlying implementation to the controller. Somebody might choose to make an amphora which can\u0027t deal with listeners one-by-one...\n\nBut the main argument I want to make is \"What harm is being done in sending all the configuration every time in .5\" - we will need that method anyway when we are failing over an amphora so I think it makes perfect sense to leave listeners et. al. out in .5 and implement more granular updates for 1.0.\n\n \"How do we delete a listener if we\u0027re only pushing out information at the \"loadbalancer\" level?\" -- The underlying amphora can compare here state with what it\u0027s send and figure out a listener disappeared. In the haproxy case there will jut be not a file for the haproxy we don\u0027t need any longer.\n\nI am not completely opposed to having listener methods but I think we are adding some unnecessary complexity at this stage.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"819e02c60d5dfd1f6ea24fda5073d857e04c5536","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_56c2f953","line":38,"in_reply_to":"fa98f980_bcd6bcef","updated":"2014-09-18 18:07:59.000000000","message":"I am still thinking for .5 we should just push the whole thing to simplify the interface for now. Adding listeners will put me on the slippery slope of why not nodes, healthmonitors, etc...","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"2db2804c45f37f084ed4e00499f10f302b6d5a4e","unresolved":false,"context_lines":[{"line_number":35,"context_line":"            #return the logger to use - this is a way to inject a custom logger for testing, etc"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"        def update(self, lb):"},{"line_number":38,"context_line":"            #update the amphorae with a new configuration for the load balancer, etc."},{"line_number":39,"context_line":"            raise NotImplementedError"},{"line_number":40,"context_line":""},{"line_number":41,"context_line":"        def suspend(self, lb):"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_98e9e4b9","line":38,"in_reply_to":"fa98f980_bed4dbbb","updated":"2014-09-19 03:34:42.000000000","message":"What other blueprint are you talking about? The one that implements the REST interface that talks with the Amphora?\n\nFor the record, we already discussed this: Updating both listeners gets you effectively the same behavior as having 1 haproxy process per load balancer, which was voted against-- one of the reasons for which was fault isolation. While we can still do 1 haproxy process per listener and get fault isolation during normal running conditions even if we\u0027re just using \"loadbalancer\"-granularity methods in this class, by unnecessarily pushing out haproxy configs / restarting processes for listeners that aren\u0027t changing you lose fault isolation when configs are being changed. I\u0027m very much against this.\n\nNow, if we are pushing out separate configs for each listener each time, we can still avoid unnecessarily restarting the haproxy process for unchanged listeners by writing code on the amphora which over-rides a restart if the new haproxy config is exactly the same as the old one (ie. lazy restarts). This pushes a bit of intelligence out to the amphora that doesn\u0027t necessarily need to be there, and also means we\u0027re effectively relying on some state information on the amphora, which I would prefer to avoid but won\u0027t veto, because this process would be idempotent. I would prefer this kind of intelligence to live in the controller, because it\u0027s more central-- but understand that we can\u0027t dogmatically follow that \"centralize intelligence\" ideal without making the overall system suffer. (Just note that sometimes, it may make sense for the controller to effectively force an haproxy process restart even without changing the haproxy config-- meaning we\u0027ll have to have listener methods to do this.)\n\nI don\u0027t see any \"slippery slope\" here. Does anyone have a valid reason for wanting to be able to push out just a healthmonitor config, just a member config, etc.?\n\nIf we\u0027re not going to allow the driver to have direct DB access, then the smallest amount of information we can hand to the driver when updating something as far down the tree as (for example) a member is \"everything necessary for generating a complete haproxy config\" (unless we\u0027re forcing the driver to keep state on the amphora or something so it doesn\u0027t have to go to the DB for this information, which is a Very Bad Idea.) A single haproxy configuration corresponds with a Listener, not a Loadbalancer.\n\nFor that matter, about the only things I see happening at the loadbalancer level are those things necessary for the amphora to bind to a virtual IP address. :P  When it comes to pushing out listener configuration and control, these should be different methods, IMO.\n\nOne other as yet unanswered question: How do we delete a listener if we\u0027re only pushing out information at the \"loadbalancer\" level?\n\nAnyway... more on this below...","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"2db2804c45f37f084ed4e00499f10f302b6d5a4e","unresolved":false,"context_lines":[{"line_number":56,"context_line":"            #some information might come from querying the amphora"},{"line_number":57,"context_line":"            raise NotImplementedError"},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"        def archive_logs(self, loadbalancer, listener\u003dALL, target\u003dSWIFT):"},{"line_number":60,"context_line":"            #archives logs for a specific load balancer, listener, etc."},{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_18dfb4d7","line":59,"updated":"2014-09-19 03:34:42.000000000","message":"If we\u0027re going to leave this in right now, I\u0027d like to see some logging requirements. Otherwise, what are we implementing? A NotImplementedError?\n\n(Of course we\u0027re going to need logging at some point-- but let\u0027s do that in another blueprint / spec, eh.)","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"73c0a71ba319aa2e7e766b90d1e9939004681ec9","unresolved":false,"context_lines":[{"line_number":56,"context_line":"            #some information might come from querying the amphora"},{"line_number":57,"context_line":"            raise NotImplementedError"},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"        def archive_logs(self, loadbalancer, listener\u003dALL, target\u003dSWIFT):"},{"line_number":60,"context_line":"            #archives logs for a specific load balancer, listener, etc."},{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_c0a0fff6","line":59,"in_reply_to":"fa98f980_18dfb4d7","updated":"2014-09-22 21:59:27.000000000","message":"Ok, I will delete logging.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"24eb1a06a05261de5e369bfed1a76b54bdbb35b9","unresolved":false,"context_lines":[{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"        def get_metrics(self, amphora):"},{"line_number":64,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them"},{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_1cb3480b","line":64,"updated":"2014-09-18 05:14:19.000000000","message":"German and I had a conversation earlier that some may not want/able to send the metrics to ceilometer (RAX being one), so whatever this calls to actually send the data off needs to be yet another driver/plugin based off an interface.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d97047174626cb37b5d844888026dfc618c4a8a","unresolved":false,"context_lines":[{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"        def get_metrics(self, amphora):"},{"line_number":64,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them"},{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_4f419759","line":64,"in_reply_to":"da9df570_808d177e","updated":"2014-09-23 03:44:16.000000000","message":"Ok, are these metrics per load balancer, per listener, per pool, etc.? Let\u0027s see methods added for the corresponding Neutron LBaaS v2 metrics.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"73c0a71ba319aa2e7e766b90d1e9939004681ec9","unresolved":false,"context_lines":[{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"        def get_metrics(self, amphora):"},{"line_number":64,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them"},{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_808d177e","line":64,"in_reply_to":"fa98f980_14b76af6","updated":"2014-09-22 21:59:27.000000000","message":"Well, we have to deliver metrics per the Neutron LBaaS v2 driver - so at least we need to have enough metrics to fulfill that spec which I think is a good starting point. So there is some requirement...","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"2db2804c45f37f084ed4e00499f10f302b6d5a4e","unresolved":false,"context_lines":[{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"        def get_metrics(self, amphora):"},{"line_number":64,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them"},{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_58977ceb","line":64,"in_reply_to":"fa98f980_1cb3480b","updated":"2014-09-19 03:34:42.000000000","message":"Right... so for now, let\u0027s leave this off and implement metrics gathering in another blueprint / spec (which will probably have to land before v0.5 is final).","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"4ed4701311e8728cb0b770dc0856db74105c0c90","unresolved":false,"context_lines":[{"line_number":61,"context_line":"            raise NotImplementedError"},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"        def get_metrics(self, amphora):"},{"line_number":64,"context_line":"            #return ceilometer ready metrics - some amphora might choose to send them"},{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_14b76af6","line":64,"in_reply_to":"fa98f980_58977ceb","updated":"2014-09-19 18:34:34.000000000","message":"Fine by me.  I prefer to work iteratively anyway.  Get a good foundation in first, then build off that.  I would like that blueprint to be created immediately after this has been decided.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"2db2804c45f37f084ed4e00499f10f302b6d5a4e","unresolved":false,"context_lines":[{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""},{"line_number":68,"context_line":"        def get_health(self, amphora):"},{"line_number":69,"context_line":"            #returns map: {\"amphora-status\":HEALTHY, loadbalancers: {\"loadbalancer-id\": {\"loadbalancer-status\": HEALTHY,"},{"line_number":70,"context_line":"            # \"listeners\":{\"listener-id\":{\"listener-status\":HEALTHY, \"nodes\":{\"node-id\":HEALTHY, ...}}, ...}, ...}}"},{"line_number":71,"context_line":"            raise NotImplementedError"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_38171856","line":68,"updated":"2014-09-19 03:34:42.000000000","message":"Ok, so having thought more about this, I\u0027m OK with a generic \u0027get_health\u0027 method that gets the whole kit and kaboodle of health information on the amphora.\n\nThe scenario I was worried about here is a user navigating through a web UI which is \"helpfully\" showing her the health on all her loadbalancers across all amphorae and their tree of child objects with every click. If that health information is expensive to gather you could easily cause problems for the amphora just by navigating through such an interface too quickly. I suppose, though, that if that does become a problem we could implement some form of caching (probably for a short interval on the amphora itself). So... my worry might otherwise have lead to some premature optimization.\n\nCarry on, eh!","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"73c0a71ba319aa2e7e766b90d1e9939004681ec9","unresolved":false,"context_lines":[{"line_number":65,"context_line":"            #straight to ceilometer others might use the mixin"},{"line_number":66,"context_line":"            raise NotImplementedError"},{"line_number":67,"context_line":""},{"line_number":68,"context_line":"        def get_health(self, amphora):"},{"line_number":69,"context_line":"            #returns map: {\"amphora-status\":HEALTHY, loadbalancers: {\"loadbalancer-id\": {\"loadbalancer-status\": HEALTHY,"},{"line_number":70,"context_line":"            # \"listeners\":{\"listener-id\":{\"listener-status\":HEALTHY, \"nodes\":{\"node-id\":HEALTHY, ...}}, ...}, ...}}"},{"line_number":71,"context_line":"            raise NotImplementedError"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_a0e93b1c","line":68,"in_reply_to":"fa98f980_38171856","updated":"2014-09-22 21:59:27.000000000","message":"+1","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"2db2804c45f37f084ed4e00499f10f302b6d5a4e","unresolved":false,"context_lines":[{"line_number":77,"context_line":"            raise NotImplementedError"},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"The referenced lb is a load balancer object as described in our model."},{"line_number":80,"context_line":"The model is detached from the DB so the driver"},{"line_number":81,"context_line":"can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":82,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at"},{"line_number":83,"context_line":"this juncture. This might be added in"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_737b817b","line":80,"updated":"2014-09-19 03:34:42.000000000","message":"Continuing the discussion from the last revision: Again, if we\u0027re not allowing the drivers to have DB access (which is a good idea, I agree), then we\u0027ll have to push out everything the driver needs to know to generate complete haproxy configs each time.\n\nWe can send the driver the whole loadbalancer config when there\u0027s an update or addition and rely on the amphora to detect when a haproxy.cfg hasn\u0027t changed for a given listener (ie. lazy restart). I\u0027m actually OK with this (because you will need to know the loadbalancer\u0027s IP addresses in the haproxy config anyway)... but then we\u0027re still going to need (at the very least) methods for starting, stopping and deleting a listener (to over-ride the lazy-restart behavior that gets us fault isolation when configs are pushed).\n\nAlternatively it might be cleaner to just make methods which operate on the listener (and embed the loadbalancer VIP information in the data structure describing the listener). Doing this, it might make sense to continue to have loadbalancer methods (mostly for binding / unbinding to IP addresses), though we should probably try to not repeat listener-manipulation functionality in the loadbalancer methods.\n\nDoes that make sense? What do you agree / disagree with above?","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d97047174626cb37b5d844888026dfc618c4a8a","unresolved":false,"context_lines":[{"line_number":77,"context_line":"            raise NotImplementedError"},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"The referenced lb is a load balancer object as described in our model."},{"line_number":80,"context_line":"The model is detached from the DB so the driver"},{"line_number":81,"context_line":"can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":82,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at"},{"line_number":83,"context_line":"this juncture. This might be added in"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_0f7f8fe1","line":80,"in_reply_to":"da9df570_c0639faf","updated":"2014-09-23 03:44:16.000000000","message":"I don\u0027t think Brandon is saying we\u0027d make changes to the data structure. I think he\u0027s saying that we pass the relevant VIP objects along with the listener object (as a secondary or tertiary argument) when listener methods are invoked in the driver.\n\nGerman: You keep saying that it\u0027s more complex to use listener methods rather than load balancer methods. Why? I don\u0027t agree this is the case, and would like to hear your reasoning for this.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"ef3a10f448dc85b475b9ae03be1ad3585f93b00d","unresolved":false,"context_lines":[{"line_number":77,"context_line":"            raise NotImplementedError"},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"The referenced lb is a load balancer object as described in our model."},{"line_number":80,"context_line":"The model is detached from the DB so the driver"},{"line_number":81,"context_line":"can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":82,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at"},{"line_number":83,"context_line":"this juncture. This might be added in"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_0a7a1d88","line":80,"in_reply_to":"da9df570_c0639faf","updated":"2014-09-23 03:31:56.000000000","message":"That data structure already exists. Would be pretty simple.\nAnyway, in the interest of getting this off the ground, i don\u0027t think it will terribly bad if we have to add in the listener methods later if it is determined that we really need them.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"4ed4701311e8728cb0b770dc0856db74105c0c90","unresolved":false,"context_lines":[{"line_number":77,"context_line":"            raise NotImplementedError"},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"The referenced lb is a load balancer object as described in our model."},{"line_number":80,"context_line":"The model is detached from the DB so the driver"},{"line_number":81,"context_line":"can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":82,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at"},{"line_number":83,"context_line":"this juncture. This might be added in"}],"source_content_type":"text/x-rst","patch_set":2,"id":"fa98f980_b4d21e32","line":80,"in_reply_to":"fa98f980_737b817b","updated":"2014-09-19 18:34:34.000000000","message":"another alternative is to pass the vip object as another parameter to the listener methods.  I do believe listener methods in this driver are the best solution so far.\n\nAlso, I\u0027ve made a decision in the API that a user\u0027s load balancer\u0027s vip cannot be altered after creation.  Not sure if this is something we want to keep but it seemed the safest thing at first.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"6d97047174626cb37b5d844888026dfc618c4a8a","unresolved":false,"context_lines":[{"line_number":77,"context_line":"            raise NotImplementedError"},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"The referenced lb is a load balancer object as described in our model."},{"line_number":80,"context_line":"The model is detached from the DB so the driver"},{"line_number":81,"context_line":"can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":82,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at"},{"line_number":83,"context_line":"this juncture. This might be added in"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_0f186fd2","line":80,"in_reply_to":"fa98f980_b4d21e32","updated":"2014-09-23 03:44:16.000000000","message":"I\u0027m OK with the load balancer\u0027s VIP being un-alterable after creation (I don\u0027t see a whole lot of need for this anyway). I could see us wanting to add a VIP to a load balancer if we eventually allow multiple VIPs per load balancer-- but we can cross that bridge when we get there.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"73c0a71ba319aa2e7e766b90d1e9939004681ec9","unresolved":false,"context_lines":[{"line_number":77,"context_line":"            raise NotImplementedError"},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"The referenced lb is a load balancer object as described in our model."},{"line_number":80,"context_line":"The model is detached from the DB so the driver"},{"line_number":81,"context_line":"can\u0027t write to the DB. Because our initial goal is to render a whole config"},{"line_number":82,"context_line":"no special methods for adding nodes, health monitors, etc. are supported at"},{"line_number":83,"context_line":"this juncture. This might be added in"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da9df570_c0639faf","line":80,"in_reply_to":"fa98f980_b4d21e32","updated":"2014-09-22 21:59:27.000000000","message":"That we need to make data structure changes tell me we should just ship complete lb\u0027s in .5 and save that for 1.0. But in any case we can define our own data structures we ship to the driver which might/might not be a copy of what\u0027s in the DB. I just wouldn\u0027t advocate that for .5 since it add more complexity I like to avoid.","commit_id":"93c628d1304e4c3016b161e99f533340e6e98dfa"}]}
