)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":9816,"name":"Takashi Kajinami","email":"kajinamit@oss.nttdata.com","username":"kajinamit"},"change_message_id":"f977d1c502948c497a98ea781d95ba22d87273ac","unresolved":false,"context_lines":[{"line_number":9,"context_line":"Currently memcache only listens to the internal_api IP. We want to"},{"line_number":10,"context_line":"make it listen to both localhost and internal_api IP because in the"},{"line_number":11,"context_line":"future some services on some roles may want to just use the localhost"},{"line_number":12,"context_line":"memcached instance instead of the using multiple memcached servers"},{"line_number":13,"context_line":"over the network as it is the case currently."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Change-Id: I6ea989ae546ef344a9b3d2dd6c86a49822efc6e2"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"3f65232a_882efcf7","line":13,"range":{"start_line":12,"start_character":19,"end_line":13,"end_character":45},"updated":"2020-10-27 16:55:38.000000000","message":"I\u0027m curious to know what would be the motivation to use local memcached instead of clustered one ? (maybe stability issue I guess)\n\nIf we use only local one then it means that we expect more often cache miss until the client hit the all controller nodes. So from perspective of performance it doesn\u0027t sound like optimal option.","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"},{"author":{"_account_id":9816,"name":"Takashi Kajinami","email":"kajinamit@oss.nttdata.com","username":"kajinamit"},"change_message_id":"6db1d1ee46ee9d027edeb8b0da65c9cd7d7f9af5","unresolved":false,"context_lines":[{"line_number":9,"context_line":"Currently memcache only listens to the internal_api IP. We want to"},{"line_number":10,"context_line":"make it listen to both localhost and internal_api IP because in the"},{"line_number":11,"context_line":"future some services on some roles may want to just use the localhost"},{"line_number":12,"context_line":"memcached instance instead of the using multiple memcached servers"},{"line_number":13,"context_line":"over the network as it is the case currently."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Change-Id: I6ea989ae546ef344a9b3d2dd6c86a49822efc6e2"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"3f65232a_223420fb","line":13,"range":{"start_line":12,"start_character":19,"end_line":13,"end_character":45},"in_reply_to":"3f65232a_4e286d85","updated":"2020-10-27 23:23:25.000000000","message":"Thanks for clarification.\n\nI\u0027m interested in this because I recently implemented general memcached caching in authtoken middleware and caching in keystone as per deprecation message from authtoken middleware and discussion about the advantage of caching in keystone for quicker token verification...","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"},{"author":{"_account_id":20172,"name":"Michele Baldessari","email":"michele@acksyn.org","username":"michele"},"change_message_id":"a7bcb5e2120ab4d3421565b6a590d5e7c644e05e","unresolved":false,"context_lines":[{"line_number":9,"context_line":"Currently memcache only listens to the internal_api IP. We want to"},{"line_number":10,"context_line":"make it listen to both localhost and internal_api IP because in the"},{"line_number":11,"context_line":"future some services on some roles may want to just use the localhost"},{"line_number":12,"context_line":"memcached instance instead of the using multiple memcached servers"},{"line_number":13,"context_line":"over the network as it is the case currently."},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Change-Id: I6ea989ae546ef344a9b3d2dd6c86a49822efc6e2"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":2,"id":"3f65232a_4e286d85","line":13,"range":{"start_line":12,"start_character":19,"end_line":13,"end_character":45},"in_reply_to":"3f65232a_882efcf7","updated":"2020-10-27 17:16:18.000000000","message":"It really depends on the service. The idea is that at least for nova metadata we switch to source routing (https://review.opendev.org/759052) so we always hit the same metadata worker.\nThe high-level idea is to use the localhost one on controller nodes (or on roles where the role memcached is installed and configured) and retain the internal_api memcached triplet for computes and roles which do not have it.\n\nThe rationale is that:\na) using localhost is faster and we should not rely on remote memcached being hot (in case we can improve the source lb selection for services)\nb) if a controller node goes down we behave better, since py-memcached has horrible \"node is gone\" detection and constantly times out when a service is down.\n\nThere has been a rather long internal thread about this I can forward you the details if you want.\n\nIn any case these are just preparation patches to give us the option of changing the memcached setup a little, the real switch will happen in other reviews once we get to it and we\u0027ll need to test those more before confirming.","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"}],"deployment/memcached/memcached-container-puppet.yaml":[{"author":{"_account_id":9816,"name":"Takashi Kajinami","email":"kajinamit@oss.nttdata.com","username":"kajinamit"},"change_message_id":"f977d1c502948c497a98ea781d95ba22d87273ac","unresolved":false,"context_lines":[{"line_number":121,"context_line":"            # internal_api_uri -\u003e [IP]"},{"line_number":122,"context_line":"            # internal_api_subnet - \u003e IP/CIDR"},{"line_number":123,"context_line":"            memcached::listen_ip:"},{"line_number":124,"context_line":"              - 127.0.0.1"},{"line_number":125,"context_line":"              - str_replace:"},{"line_number":126,"context_line":"                  template:"},{"line_number":127,"context_line":"                    \"%{hiera(\u0027$NETWORK\u0027)}\""}],"source_content_type":"text/x-yaml","patch_set":2,"id":"3f65232a_28a368a0","line":124,"range":{"start_line":124,"start_character":16,"end_line":124,"end_character":25},"updated":"2020-10-27 16:55:38.000000000","message":"IIRC we always have ::1 enabled in overcloud nodes, even if we disable IPv6 by the KernelDisableIPv6 parameter, so it might be better to use ::1 instead of 127.0.0.1 .\n\nAnother option would be to use MemcachedNetwork to check which IP stack it should use, as is done in redis deployment templates.","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"},{"author":{"_account_id":20172,"name":"Michele Baldessari","email":"michele@acksyn.org","username":"michele"},"change_message_id":"a7bcb5e2120ab4d3421565b6a590d5e7c644e05e","unresolved":false,"context_lines":[{"line_number":121,"context_line":"            # internal_api_uri -\u003e [IP]"},{"line_number":122,"context_line":"            # internal_api_subnet - \u003e IP/CIDR"},{"line_number":123,"context_line":"            memcached::listen_ip:"},{"line_number":124,"context_line":"              - 127.0.0.1"},{"line_number":125,"context_line":"              - str_replace:"},{"line_number":126,"context_line":"                  template:"},{"line_number":127,"context_line":"                    \"%{hiera(\u0027$NETWORK\u0027)}\""}],"source_content_type":"text/x-yaml","patch_set":2,"id":"3f65232a_8ea56507","line":124,"range":{"start_line":124,"start_character":16,"end_line":124,"end_character":25},"in_reply_to":"3f65232a_28a368a0","updated":"2020-10-27 17:16:18.000000000","message":"We\u0027re kind of trying to get away from users having to specify RedisIPv6 and the likes as it is error prone and we should do better/simpler in general.\n\nIn this case I believe it is more likely that 127.0.0.1 always exists rather than ::1 so it seems the safer option in general?","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"},{"author":{"_account_id":9816,"name":"Takashi Kajinami","email":"kajinamit@oss.nttdata.com","username":"kajinamit"},"change_message_id":"6db1d1ee46ee9d027edeb8b0da65c9cd7d7f9af5","unresolved":false,"context_lines":[{"line_number":121,"context_line":"            # internal_api_uri -\u003e [IP]"},{"line_number":122,"context_line":"            # internal_api_subnet - \u003e IP/CIDR"},{"line_number":123,"context_line":"            memcached::listen_ip:"},{"line_number":124,"context_line":"              - 127.0.0.1"},{"line_number":125,"context_line":"              - str_replace:"},{"line_number":126,"context_line":"                  template:"},{"line_number":127,"context_line":"                    \"%{hiera(\u0027$NETWORK\u0027)}\""}],"source_content_type":"text/x-yaml","patch_set":2,"id":"3f65232a_e21ce8c2","line":124,"range":{"start_line":124,"start_character":16,"end_line":124,"end_character":25},"in_reply_to":"3f65232a_8ea56507","updated":"2020-10-27 23:23:25.000000000","message":"I didn\u0027t intend to add a new parameter like MemcachedIPv6 but check the ip version assigned to MemcachedNetwork as we do for redis with RedisNetwork.\n\nThe availability of 127.0.0.1 might depend on the way how we implemente IPv6 only deployment, but since I don\u0027t see any implementation implying that 127.0.0.1 is disabled, I think we are good to use 127.0.0.1 .","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"},{"author":{"_account_id":20172,"name":"Michele Baldessari","email":"michele@acksyn.org","username":"michele"},"change_message_id":"d554e843042286bc95c84c39c08bb18044a296bb","unresolved":false,"context_lines":[{"line_number":121,"context_line":"            # internal_api_uri -\u003e [IP]"},{"line_number":122,"context_line":"            # internal_api_subnet - \u003e IP/CIDR"},{"line_number":123,"context_line":"            memcached::listen_ip:"},{"line_number":124,"context_line":"              - 127.0.0.1"},{"line_number":125,"context_line":"              - str_replace:"},{"line_number":126,"context_line":"                  template:"},{"line_number":127,"context_line":"                    \"%{hiera(\u0027$NETWORK\u0027)}\""}],"source_content_type":"text/x-yaml","patch_set":2,"id":"1f621f24_7613ce67","line":124,"range":{"start_line":124,"start_character":16,"end_line":124,"end_character":25},"in_reply_to":"3f65232a_e21ce8c2","updated":"2020-10-28 08:48:22.000000000","message":"Ah I see now, thanks. That makes sense I\u0027ll amend it so we\u0027re consistent across services","commit_id":"8b998758ac494dee1b4cfaefb7fb210e9e3bde40"}]}