)]}'
{"specs/kilo/swift-retry-wait.rst":[{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"3a3edcbef302b4bea1121c25c76b785855b28659","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Firstly,"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":".. code:: python"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"    swift_store_retry_get_wait \u003d 0.2"},{"line_number":38,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_42333862","line":35,"updated":"2015-01-12 15:56:30.000000000","message":"Isn\u0027t it code-block? I think .. code:: doesn\u0027t allow you to specify a language. (Same below)\n\nFurther, should this be python or ini?","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ac1fa9ab9d9c52b6b0a97dd215cced03fc93395c","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Firstly,"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":".. code:: python"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"    swift_store_retry_get_wait \u003d 0.2"},{"line_number":38,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_65383e3e","line":35,"in_reply_to":"3a961159_42333862","updated":"2015-01-12 16:09:59.000000000","message":"I\u0027ve always used .. code::, but although it works in the platforms I\u0027ve used, it  looks like a docutils specific extension. I\u0027ll change this.","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"3a3edcbef302b4bea1121c25c76b785855b28659","unresolved":false,"context_lines":[{"line_number":49,"context_line":"If both ``swift_store_retry_get_count`` and ``swift_store_retry_get_wait`` are"},{"line_number":50,"context_line":"non-zero, setting this to True causes each successive retry delay to be"},{"line_number":51,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":52,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_2502b609","line":52,"updated":"2015-01-12 15:56:30.000000000","message":"Is there a cap on how many times we will retry? Should there be?","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ac1fa9ab9d9c52b6b0a97dd215cced03fc93395c","unresolved":false,"context_lines":[{"line_number":49,"context_line":"If both ``swift_store_retry_get_count`` and ``swift_store_retry_get_wait`` are"},{"line_number":50,"context_line":"non-zero, setting this to True causes each successive retry delay to be"},{"line_number":51,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":52,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_e5a7cea5","line":52,"in_reply_to":"3a961159_2502b609","updated":"2015-01-12 16:09:59.000000000","message":"The cap is set with swift_store_retry_get_count, which already exists.","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"d0cf00cee79564d95fc194623b7939d3944c26f2","unresolved":false,"context_lines":[{"line_number":49,"context_line":"If both ``swift_store_retry_get_count`` and ``swift_store_retry_get_wait`` are"},{"line_number":50,"context_line":"non-zero, setting this to True causes each successive retry delay to be"},{"line_number":51,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":52,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_485519f9","line":52,"in_reply_to":"3a961159_e5a7cea5","updated":"2015-01-12 16:20:47.000000000","message":"Ah, I missed that above. Sorry.","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"3a3edcbef302b4bea1121c25c76b785855b28659","unresolved":false,"context_lines":[{"line_number":61,"context_line":"using a \u0027one size fits all\u0027 exponent of 2."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"Having a simple on/off for exponential backoff is easier to understand and"},{"line_number":64,"context_line":"configure, however."},{"line_number":65,"context_line":""},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Data model impact"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_2529d689","line":64,"updated":"2015-01-12 15:56:30.000000000","message":"Editorial nit: Lose the \", however\".","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"d698dc4986daa75ab8fd58f4e6a79d35ebd73950","unresolved":false,"context_lines":[{"line_number":61,"context_line":"using a \u0027one size fits all\u0027 exponent of 2."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"Having a simple on/off for exponential backoff is easier to understand and"},{"line_number":64,"context_line":"configure, however."},{"line_number":65,"context_line":""},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"Data model impact"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_28bd2dd8","line":64,"in_reply_to":"3a961159_2529d689","updated":"2015-01-12 16:28:41.000000000","message":"Done","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"3a3edcbef302b4bea1121c25c76b785855b28659","unresolved":false,"context_lines":[{"line_number":129,"context_line":"----------"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"- add new sleep behaviour"},{"line_number":132,"context_line":"- add options to control this behaviour"},{"line_number":133,"context_line":"- add tests"},{"line_number":134,"context_line":""},{"line_number":135,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_c531f2ee","line":132,"updated":"2015-01-12 15:56:30.000000000","message":"document the options too ;) Technically this is already done, but it\u0027s good to mention it here anyway.","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"d698dc4986daa75ab8fd58f4e6a79d35ebd73950","unresolved":false,"context_lines":[{"line_number":129,"context_line":"----------"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"- add new sleep behaviour"},{"line_number":132,"context_line":"- add options to control this behaviour"},{"line_number":133,"context_line":"- add tests"},{"line_number":134,"context_line":""},{"line_number":135,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_48b27907","line":132,"in_reply_to":"3a961159_c531f2ee","updated":"2015-01-12 16:28:41.000000000","message":"Done","commit_id":"637062f867748be90e3326eff2a67180a9c7212b"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"92cdbdde77883ff04c2d3ee0129fbbda5f164b0e","unresolved":false,"context_lines":[{"line_number":17,"context_line":"Problem description"},{"line_number":18,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":19,"context_line":""},{"line_number":20,"context_line":"If getting an image from swift fails, it is probable that the failure will"},{"line_number":21,"context_line":"occur again if the request is retried immediately if the underlying issue is"},{"line_number":22,"context_line":"network connectivity or a hardware failure."},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_37fca472","line":20,"updated":"2015-01-12 19:42:53.000000000","message":"After a failed image-data download from swift  ...\n\nit is more likely that the failure will","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"62a5b7086a6a85d2df76e57b37bb08ef1cb9d14b","unresolved":false,"context_lines":[{"line_number":17,"context_line":"Problem description"},{"line_number":18,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":19,"context_line":""},{"line_number":20,"context_line":"If getting an image from swift fails, it is probable that the failure will"},{"line_number":21,"context_line":"occur again if the request is retried immediately if the underlying issue is"},{"line_number":22,"context_line":"network connectivity or a hardware failure."},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_ff7bd2f1","line":20,"in_reply_to":"3a961159_37fca472","updated":"2015-01-12 20:25:19.000000000","message":"+1 on the wording change.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ae34f2fe32f98e76e87939f4b410138604f427b1","unresolved":false,"context_lines":[{"line_number":17,"context_line":"Problem description"},{"line_number":18,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":19,"context_line":""},{"line_number":20,"context_line":"If getting an image from swift fails, it is probable that the failure will"},{"line_number":21,"context_line":"occur again if the request is retried immediately if the underlying issue is"},{"line_number":22,"context_line":"network connectivity or a hardware failure."},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_e27179ac","line":20,"in_reply_to":"3a961159_ff7bd2f1","updated":"2015-01-12 20:42:20.000000000","message":"Done","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":5314,"name":"Brian Rosmaita","email":"rosmaita.fossdev@gmail.com","username":"brian-rosmaita"},"change_message_id":"35827b666c18a683817bed3b084c1311865931a4","unresolved":false,"context_lines":[{"line_number":34,"context_line":""},{"line_number":35,"context_line":".. code-block:: ini"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"    swift_store_retry_get_wait \u003d 0.2"},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"This specifies the number of seconds delayed between each Swift store get"},{"line_number":40,"context_line":"requests. Enabling this will only have any effect if"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_f7b93c7d","line":37,"updated":"2015-01-12 19:01:30.000000000","message":"I don\u0027t think we have any other config options that are real numbers ... maybe make this milliseconds?  (1) This is just an observation, I don\u0027t insist on it. (2) since this is the first float in the config file, it\u0027s worth thinking about.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"6820864a2884ed163bec81d00cc084d109459a13","unresolved":false,"context_lines":[{"line_number":34,"context_line":""},{"line_number":35,"context_line":".. code-block:: ini"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"    swift_store_retry_get_wait \u003d 0.2"},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"This specifies the number of seconds delayed between each Swift store get"},{"line_number":40,"context_line":"requests. Enabling this will only have any effect if"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_0998701a","line":37,"in_reply_to":"3a961159_f7b93c7d","updated":"2015-01-12 19:25:31.000000000","message":"I was originally going with milliseconds, but couldn\u0027t see any usage of them in the glance config. I\u0027m okay with either, so if you think they would fit better, I can change it back to milliseconds.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"92cdbdde77883ff04c2d3ee0129fbbda5f164b0e","unresolved":false,"context_lines":[{"line_number":87,"context_line":"Other end user impact"},{"line_number":88,"context_line":"---------------------"},{"line_number":89,"context_line":""},{"line_number":90,"context_line":"This will result in fewer failures to retrieve an image from glance in certain"},{"line_number":91,"context_line":"cases."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_890d4038","line":90,"updated":"2015-01-12 19:42:53.000000000","message":"to retrieve the image data","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ae34f2fe32f98e76e87939f4b410138604f427b1","unresolved":false,"context_lines":[{"line_number":87,"context_line":"Other end user impact"},{"line_number":88,"context_line":"---------------------"},{"line_number":89,"context_line":""},{"line_number":90,"context_line":"This will result in fewer failures to retrieve an image from glance in certain"},{"line_number":91,"context_line":"cases."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_42716da7","line":90,"in_reply_to":"3a961159_7f67e251","updated":"2015-01-12 20:42:20.000000000","message":"Done","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"62a5b7086a6a85d2df76e57b37bb08ef1cb9d14b","unresolved":false,"context_lines":[{"line_number":87,"context_line":"Other end user impact"},{"line_number":88,"context_line":"---------------------"},{"line_number":89,"context_line":""},{"line_number":90,"context_line":"This will result in fewer failures to retrieve an image from glance in certain"},{"line_number":91,"context_line":"cases."},{"line_number":92,"context_line":""},{"line_number":93,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_7f67e251","line":90,"in_reply_to":"3a961159_890d4038","updated":"2015-01-12 20:25:19.000000000","message":"+1","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"92cdbdde77883ff04c2d3ee0129fbbda5f164b0e","unresolved":false,"context_lines":[{"line_number":94,"context_line":"Performance Impact"},{"line_number":95,"context_line":"------------------"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"None"},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"Other deployer impact"},{"line_number":100,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_09631042","line":97,"updated":"2015-01-12 19:42:53.000000000","message":"My understanding is that this _may_ have performance impact on the network connections on the node as well as the speed of the download based on the values set. Might be a good place to give that disclaimer.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"62a5b7086a6a85d2df76e57b37bb08ef1cb9d14b","unresolved":false,"context_lines":[{"line_number":94,"context_line":"Performance Impact"},{"line_number":95,"context_line":"------------------"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"None"},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"Other deployer impact"},{"line_number":100,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_dfc45610","line":97,"in_reply_to":"3a961159_09631042","updated":"2015-01-12 20:25:19.000000000","message":"+1. At least it may increase the bandwidth used by the image service nodes. It may not affect actual network performance but it /could/.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"0d19b0e9828019d3095b10bf491ec7c4616bda3c","unresolved":false,"context_lines":[{"line_number":94,"context_line":"Performance Impact"},{"line_number":95,"context_line":"------------------"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"None"},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"Other deployer impact"},{"line_number":100,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_265a1582","line":97,"in_reply_to":"3a961159_0c82a5fe","updated":"2015-01-16 15:59:04.000000000","message":"Thanks for getting back on this, Stuart.\n\nI believe we are on the same page. Although such a change by itself would not be a reason for _new_ failure mode, it can be confusing for unfamiliar operators.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":455,"name":"Stuart McLaren","email":"stuart.mclaren@hpe.com","username":"stuart-mclaren"},"change_message_id":"386bcd3241cd39caf2176177f5347823a92ad05f","unresolved":false,"context_lines":[{"line_number":94,"context_line":"Performance Impact"},{"line_number":95,"context_line":"------------------"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"None"},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"Other deployer impact"},{"line_number":100,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_0c82a5fe","line":97,"in_reply_to":"3a961159_61418ea5","updated":"2015-01-16 10:45:16.000000000","message":"@nikhil: You bring up a good point around timeouts.\n\nAs an implementation detail we should probably mention in the help that other timeout values should be considered when setting these new config values.\n\nI *think* that we are not introducing a new failure mode per se, it is already possible to get a timeout between the glance client \u003c-\u003e networking h/w \u003c-\u003e glance api server for various reasons; but it\u0027s true that the possibility of hitting such a timeout may be increased with very high swift retry times.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"d972f85886e2fa5c70632707eba9abaa31b7768e","unresolved":false,"context_lines":[{"line_number":94,"context_line":"Performance Impact"},{"line_number":95,"context_line":"------------------"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"None"},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"Other deployer impact"},{"line_number":100,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_61418ea5","line":97,"in_reply_to":"3a961159_62dae97a","updated":"2015-01-15 16:22:54.000000000","message":"Agreed. The b/w should not be affected. However, let\u0027s note here for bookkeeping sake that an option like exponential back off above Layer 3 can be tricky to manage. \n\nThis combined with setting for socket timeout as well as underlying settings like QoS can result into 5xx errors, leaving clients and the operators clueless.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ae34f2fe32f98e76e87939f4b410138604f427b1","unresolved":false,"context_lines":[{"line_number":94,"context_line":"Performance Impact"},{"line_number":95,"context_line":"------------------"},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"None"},{"line_number":98,"context_line":""},{"line_number":99,"context_line":"Other deployer impact"},{"line_number":100,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_62dae97a","line":97,"in_reply_to":"3a961159_dfc45610","updated":"2015-01-12 20:42:20.000000000","message":"This should not have an effect on the amount of bandwidth used, since the swift retry will resume from where it left off. However, the overall time from the image data copying could be effected badly by bad settings and a connection to swift which is interrupted regularly.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"92cdbdde77883ff04c2d3ee0129fbbda5f164b0e","unresolved":false,"context_lines":[{"line_number":100,"context_line":"---------------------"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"This change can be enabled by deployers. It is enabled if"},{"line_number":103,"context_line":"``swift_store_retry_get_count`` is also enabled (currently defaults to"},{"line_number":104,"context_line":"disabled)."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_894e20c1","line":103,"updated":"2015-01-12 19:42:53.000000000","message":"Can you please give an example here to state how to enable this?","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"62a5b7086a6a85d2df76e57b37bb08ef1cb9d14b","unresolved":false,"context_lines":[{"line_number":100,"context_line":"---------------------"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"This change can be enabled by deployers. It is enabled if"},{"line_number":103,"context_line":"``swift_store_retry_get_count`` is also enabled (currently defaults to"},{"line_number":104,"context_line":"disabled)."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_dfeb7679","line":103,"in_reply_to":"3a961159_894e20c1","updated":"2015-01-12 20:25:19.000000000","message":"+1. An example (or two) above with all of the config options necessary to take advantage of these two new options would be helpful. They could probably be included in the docs as well.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ae34f2fe32f98e76e87939f4b410138604f427b1","unresolved":false,"context_lines":[{"line_number":100,"context_line":"---------------------"},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"This change can be enabled by deployers. It is enabled if"},{"line_number":103,"context_line":"``swift_store_retry_get_count`` is also enabled (currently defaults to"},{"line_number":104,"context_line":"disabled)."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_023a0598","line":103,"in_reply_to":"3a961159_dfeb7679","updated":"2015-01-12 20:42:20.000000000","message":"I\u0027ll add a fuller example in the next PS.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":2537,"name":"Nikhil Komawar","email":"nik.komawar@gmail.com","username":"nikhil-komawar"},"change_message_id":"92cdbdde77883ff04c2d3ee0129fbbda5f164b0e","unresolved":false,"context_lines":[{"line_number":126,"context_line":"  stuart-mclaren"},{"line_number":127,"context_line":""},{"line_number":128,"context_line":"Work Items"},{"line_number":129,"context_line":"----------"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"- Add new sleep behaviour"},{"line_number":132,"context_line":"- Add options to control this behaviour"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_49b99878","line":129,"updated":"2015-01-12 19:42:53.000000000","message":"Probably good place to state what the defaults you are planning to set in the code.","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"ae34f2fe32f98e76e87939f4b410138604f427b1","unresolved":false,"context_lines":[{"line_number":126,"context_line":"  stuart-mclaren"},{"line_number":127,"context_line":""},{"line_number":128,"context_line":"Work Items"},{"line_number":129,"context_line":"----------"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"- Add new sleep behaviour"},{"line_number":132,"context_line":"- Add options to control this behaviour"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_42f6cdfa","line":129,"in_reply_to":"3a961159_49b99878","updated":"2015-01-12 20:42:20.000000000","message":"Done","commit_id":"d3218cfb67ed04d2eff66951dd8a0cc42b6b081f"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"2951519e48a52329e93d1b7ffe3554babe1f79aa","unresolved":false,"context_lines":[{"line_number":106,"context_line":"disabled)."},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"This can be enabled by setting ``swift_store_retry_get_count`` to a positive"},{"line_number":109,"context_line":"integer."},{"line_number":110,"context_line":""},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"Developer impact"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3a961159_644e7270","line":109,"updated":"2015-01-12 23:26:52.000000000","message":"Perhaps add something like:\n\n\"An example configuration that enables retries with a wait and exponential wait would look like:\n\n    swift_store_retry_get_count \u003d 5\n    swift_store_retry_get_wait \u003d 0.2\n    swift_store_retry_get_exponential_wait \u003d True\n\n\"","commit_id":"79b8cfbd76e91773fccd08ca24aa6cbde077992e"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"8e8d26ae6b0dbe0785d1fc5df3432d363a9c3b51","unresolved":false,"context_lines":[{"line_number":106,"context_line":"disabled)."},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"This can be enabled by setting ``swift_store_retry_get_count`` to a positive"},{"line_number":109,"context_line":"integer."},{"line_number":110,"context_line":""},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"Developer impact"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3a961159_7cb7d785","line":109,"in_reply_to":"3a961159_644e7270","updated":"2015-01-13 12:14:48.000000000","message":"Done","commit_id":"79b8cfbd76e91773fccd08ca24aa6cbde077992e"},{"author":{"_account_id":11864,"name":"Ben Roble","email":"ben.roble@rackspace.com","username":"broble"},"change_message_id":"1606b13bbe89b2d34249343bbffdc1c4c8a79454","unresolved":false,"context_lines":[{"line_number":114,"context_line":"    swift_store_retry_get_wait \u003d 1"},{"line_number":115,"context_line":"    swift_store_retry_get_exponential_wait \u003d True"},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":118,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"},{"line_number":119,"context_line":"``vmware_api_retry_count \u003d 10``)."},{"line_number":120,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3a961159_ff6244d7","line":117,"updated":"2015-01-16 22:03:05.000000000","message":"Since this impacts the default behavior of swift, I recommend this change also be listed under the \"Proposed change\" section in the spec for improved clarity.","commit_id":"14104164460adac7a2ab5f02d864d43600f7bb23"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"9c74cca5832278634d66ee7cdcbe9715cf9702dc","unresolved":false,"context_lines":[{"line_number":114,"context_line":"    swift_store_retry_get_wait \u003d 1"},{"line_number":115,"context_line":"    swift_store_retry_get_exponential_wait \u003d True"},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":118,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"},{"line_number":119,"context_line":"``vmware_api_retry_count \u003d 10``)."},{"line_number":120,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3a961159_a09ec253","line":117,"in_reply_to":"3a961159_dfd028f9","updated":"2015-01-19 12:06:40.000000000","message":"Done","commit_id":"14104164460adac7a2ab5f02d864d43600f7bb23"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"c9a2ec3875aeaf679b899e710b289c5701bb5922","unresolved":false,"context_lines":[{"line_number":114,"context_line":"    swift_store_retry_get_wait \u003d 1"},{"line_number":115,"context_line":"    swift_store_retry_get_exponential_wait \u003d True"},{"line_number":116,"context_line":""},{"line_number":117,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":118,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"},{"line_number":119,"context_line":"``vmware_api_retry_count \u003d 10``)."},{"line_number":120,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3a961159_dfd028f9","line":117,"in_reply_to":"3a961159_ff6244d7","updated":"2015-01-16 22:03:58.000000000","message":"Good idea.","commit_id":"14104164460adac7a2ab5f02d864d43600f7bb23"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"dfe91df5f62b4ea025bfd2eeba9179343266a5f5","unresolved":false,"context_lines":[{"line_number":51,"context_line":"non-zero, setting this boolean to True causes each successive retry delay to be"},{"line_number":52,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":53,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":54,"context_line":""},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":57,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_b8fd0e3f","line":54,"updated":"2015-01-19 17:27:46.000000000","message":"Nit: One less new-line?","commit_id":"88e719d84c093875721e5631eba21e39794125d9"},{"author":{"_account_id":11356,"name":"Louis Taylor","email":"louis@kragniz.eu","username":"kragniz"},"change_message_id":"a2962bc08630e3bfeafc7d927f6714a5c254a3a5","unresolved":false,"context_lines":[{"line_number":51,"context_line":"non-zero, setting this boolean to True causes each successive retry delay to be"},{"line_number":52,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":53,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":54,"context_line":""},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":57,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_b8ab4e1e","line":54,"in_reply_to":"3a961159_b8fd0e3f","updated":"2015-01-19 17:29:48.000000000","message":"Done","commit_id":"88e719d84c093875721e5631eba21e39794125d9"},{"author":{"_account_id":5202,"name":"Erno Kuvaja","email":"jokke@usr.fi","username":"jokke"},"change_message_id":"25c763da2982675b4ca454f4c1d5119c5a7708b1","unresolved":false,"context_lines":[{"line_number":20,"context_line":""},{"line_number":21,"context_line":"If getting an image from swift fails, it is more than likely the failure will"},{"line_number":22,"context_line":"occur again if the request is retried immediately if the underlying issue is"},{"line_number":23,"context_line":"network connectivity or a hardware failure."},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"It would be useful for a deployer to be able to specify an amount of time to"},{"line_number":26,"context_line":"wait between these retries."}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a930d6b_8610c402","line":23,"updated":"2015-01-22 12:03:06.000000000","message":"If the individual Swift node has an issue after reconnect it\u0027s unlikely to hit the same node again (unless the environment is quite small with one or so nodes). If the issue is on the Glance side we will most probably loose the client connection anyways.","commit_id":"6389c192c16858e226cdeb8327ea98e457e8b19d"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"6f68974151e2823a45448cac6195d2317d4699e5","unresolved":false,"context_lines":[{"line_number":20,"context_line":""},{"line_number":21,"context_line":"If getting an image from swift fails, it is more than likely the failure will"},{"line_number":22,"context_line":"occur again if the request is retried immediately if the underlying issue is"},{"line_number":23,"context_line":"network connectivity or a hardware failure."},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"It would be useful for a deployer to be able to specify an amount of time to"},{"line_number":26,"context_line":"wait between these retries."}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a930d6b_d10d1739","line":23,"in_reply_to":"1a930d6b_8610c402","updated":"2015-01-22 17:16:31.000000000","message":"So this is probably a little too vaguely worded. If the swift cluster is only being used by glance at that particular point in time Erno, then yes, this shouldn\u0027t be an issue. if it\u0027s a smaller cluster that\u0027s under heavy load and the node we were directed to was the least utilized and continues to be for a short while afterwards, this will be helpful. If it\u0027s network connectivity, there are ghosts in the wires. Sometimes it can be a DNS problem (e.g., DNS set up using Neutron is unavailable). If that\u0027s the case then retrying immediately won\u0027t help anyone. Backing off until the DNS service can come back up will be wonderfully helpful. These are all hypothetical situations though, not ones I\u0027ve personally experienced.","commit_id":"6389c192c16858e226cdeb8327ea98e457e8b19d"},{"author":{"_account_id":5202,"name":"Erno Kuvaja","email":"jokke@usr.fi","username":"jokke"},"change_message_id":"25c763da2982675b4ca454f4c1d5119c5a7708b1","unresolved":false,"context_lines":[{"line_number":50,"context_line":"If both ``swift_store_retry_get_count`` and ``swift_store_retry_get_wait`` are"},{"line_number":51,"context_line":"non-zero, setting this boolean to True causes each successive retry delay to be"},{"line_number":52,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":53,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":56,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a930d6b_c6f80c7f","line":53,"updated":"2015-01-22 12:03:06.000000000","message":"What is the benefit of this behavior if we close the connection and open a new one to swift anyways?\n\nIMO this introduces unnecessary delay because if we connect to new swift node it\u0027s likely not facing the issue. If we hit the same faulty node again, increasing the delay will not likely change anything. The use case for this vs. well tuned retry_get_wait and retry_get_count is not clear to me.\n\nThis would make more sense if we did this within the socket and if we loose the socket we would retry on set wait period without increments.","commit_id":"6389c192c16858e226cdeb8327ea98e457e8b19d"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"6f68974151e2823a45448cac6195d2317d4699e5","unresolved":false,"context_lines":[{"line_number":50,"context_line":"If both ``swift_store_retry_get_count`` and ``swift_store_retry_get_wait`` are"},{"line_number":51,"context_line":"non-zero, setting this boolean to True causes each successive retry delay to be"},{"line_number":52,"context_line":"doubled. For example, with an initial ``swift_store_retry_get_wait \u003d 0.5``,"},{"line_number":53,"context_line":"sleep times would be: 0.5, 1, 2, 4, 8... Etc."},{"line_number":54,"context_line":""},{"line_number":55,"context_line":"``swift_store_retry_get_count`` will be set to 5 by default. Other stores"},{"line_number":56,"context_line":"currently also have retries enabled by default (``cinder_http_retries \u003d 3``,"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a930d6b_71838356","line":53,"in_reply_to":"1a930d6b_c6f80c7f","updated":"2015-01-22 17:16:31.000000000","message":"See above why this is necessary. We can probably refine (in the future) what types of errors are retried and whether exponential back-off applies to them but I\u0027m not sure if we should. That adds a whole level of complexity and exposes way too much detail to the operator and adds a whole extra surface area for user-error when configuring glance/-store.\n\nTo reiterate: If we hit the same node again the others could be under heavier loads. Increasing the delay will help as other nodes become more available, we will eventually get a different one.\n\nI also don\u0027t know what you mean by \"well tuned\" Erno.\n\nAnd FWIW, sockets in Python don\u0027t have any retry capability inherently present except for what you add on top. So doing this \"within the socket\" doesn\u0027t make sense (to me at least).","commit_id":"6389c192c16858e226cdeb8327ea98e457e8b19d"},{"author":{"_account_id":5202,"name":"Erno Kuvaja","email":"jokke@usr.fi","username":"jokke"},"change_message_id":"25c763da2982675b4ca454f4c1d5119c5a7708b1","unresolved":false,"context_lines":[{"line_number":105,"context_line":""},{"line_number":106,"context_line":"This may cause adverse performance impact if ``swift_store_retry_get_wait`` is"},{"line_number":107,"context_line":"set to some high value on an unstable network. In this case, glance will wait a"},{"line_number":108,"context_line":"disproportionate amount of time each time the download is resumed."},{"line_number":109,"context_line":""},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"Other deployer impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a930d6b_46c89c22","line":108,"updated":"2015-01-22 12:03:06.000000000","message":"I think this concern is pretty real on exponential_wait. Even being configurable.","commit_id":"6389c192c16858e226cdeb8327ea98e457e8b19d"},{"author":{"_account_id":12000,"name":"Ian Cordasco","email":"sigmavirus24@gmail.com","username":"sigmavirus24"},"change_message_id":"6f68974151e2823a45448cac6195d2317d4699e5","unresolved":false,"context_lines":[{"line_number":105,"context_line":""},{"line_number":106,"context_line":"This may cause adverse performance impact if ``swift_store_retry_get_wait`` is"},{"line_number":107,"context_line":"set to some high value on an unstable network. In this case, glance will wait a"},{"line_number":108,"context_line":"disproportionate amount of time each time the download is resumed."},{"line_number":109,"context_line":""},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"Other deployer impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a930d6b_518cc77f","line":108,"in_reply_to":"1a930d6b_46c89c22","updated":"2015-01-22 17:16:31.000000000","message":"It\u0027s definitely something we should strive to educate operators about.","commit_id":"6389c192c16858e226cdeb8327ea98e457e8b19d"}]}
