)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":29074,"name":"Felix Huettner","email":"felix.huettner@digits.schwarz","username":"felix.huettner"},"change_message_id":"b9359ced7bf9f098e3649fb3aa1ee74ec37aad8d","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"a6c5bc68_92889778","updated":"2022-11-15 12:14:52.000000000","message":"@amorin you told me in irc that having more than one read only database probably does not make sense for load reasons. Could you maybe elaborate that here?","commit_id":"1fb7ce9cb3d7bbab6256862f4a433f04ccc4b07a"},{"author":{"_account_id":31357,"name":"LEDUC Florian","display_name":"Florian LEDUC","email":"florian.leduc@socgen.com","username":"leducflorian","status":"Société Générale"},"change_message_id":"925874e986e6874ca52adbb070f839d15bbce866","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"c405157f_ea0d98ee","in_reply_to":"0c714fa8_9b477d6f","updated":"2022-11-15 17:18:07.000000000","message":"Definitely, it\u0027s a good idea. I\u0027ll update my patch.\n\nHaving another VIP dedicated to backups with only one node enabled could be another approach.\n\nThe \"last node is the backup node\" approach could be done with setting \"weight 0\" to a node. It will never serve any requests even if the ACTIVE nodes go down. But it could still be reached through it\u0027s real IP for dumping the DB.","commit_id":"1fb7ce9cb3d7bbab6256862f4a433f04ccc4b07a"},{"author":{"_account_id":11583,"name":"Arnaud Morin","email":"arnaud.morin@gmail.com","username":"arnaudmorin"},"change_message_id":"f9e117215139d8f20f7f6a01e541c928d8c8ceb9","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"0c714fa8_9b477d6f","in_reply_to":"1c9fd5a0_f45b238b","updated":"2022-11-15 14:08:54.000000000","message":"I can explain why we dont want our last node to be read at OVH: we are using it to perform backups (using mariadbbackup).\nWhen doing a backup, the server is temporarily disabled from galera cluster (WSREP OFF) and cant reply to requests anymore.\n\nSo, one other approach to this documentation would be to document the fact that the last node is the backup node, and should never receive requests.\nThen we could keep the current config.\n\n@florian, what do you think?","commit_id":"1fb7ce9cb3d7bbab6256862f4a433f04ccc4b07a"},{"author":{"_account_id":11583,"name":"Arnaud Morin","email":"arnaud.morin@gmail.com","username":"arnaudmorin"},"change_message_id":"587a8127a6fa7bd9c71c6d4ab07bb04e46a1419b","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"1c9fd5a0_f45b238b","in_reply_to":"98763c21_d57161bd","updated":"2022-11-15 14:06:12.000000000","message":"Thinking about it, I can\u0027t see any downside of having 2 \"read\" nodes.\nWe just want to exclude the \"write\" one.\nSo patchet LGTM.","commit_id":"1fb7ce9cb3d7bbab6256862f4a433f04ccc4b07a"},{"author":{"_account_id":11583,"name":"Arnaud Morin","email":"arnaud.morin@gmail.com","username":"arnaudmorin"},"change_message_id":"92eca3cc471512522d603f63df610872bd8a4ffd","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"98763c21_d57161bd","in_reply_to":"a6c5bc68_92889778","updated":"2022-11-15 14:03:34.000000000","message":"This is a good question.\nIn our deployment we only want one server to be used for \"read\" requests, that\u0027s why I end-up with the following config.\n\nNote also that the load on the \"read\" DB is far behind the \"write\" one.\n\nI am not against the idea to have multiple read backends anyway","commit_id":"1fb7ce9cb3d7bbab6256862f4a433f04ccc4b07a"}]}