)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":22629,"name":"Michal Nasiadka","email":"mnasiadka@gmail.com","username":"mnasiadka"},"change_message_id":"7129c45c0ad4e78fd6b45c82341723c4f3a3ac31","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"438d3143_89031937","updated":"2024-10-24 15:42:11.000000000","message":"CI only and trivial","commit_id":"49cebeaf4dfaf018b06016601aa54c00cbcfa13d"},{"author":{"_account_id":27339,"name":"Michal Arbet","email":"michal.arbet@ultimum.io","username":"michalarbet"},"change_message_id":"6b414a05fb63579a596d9c02b9284ed619777e6b","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"4e57b02d_d5b33ca8","updated":"2024-10-29 10:03:29.000000000","message":"Oh, thank you very much @mnasiadka@gmail.com for this type of patch; it’s exactly what I wanted to submit. I just wanted to somehow force/bend ProxySQL to ignore a node failure for a bit longer in a single-node deployment. Unfortunately, I couldn’t manage this because, by design, it’s not an HA deployment and ProxySQL is too smart to detect this, and ProxySQL indeed sends the correct error to the client—that it can’t reach hostgroup 0 (since the node is simply down). But this is fine... because OpenStack services retry the DB connection, and everything works as expected. In a multi-node MariaDB cluster, this doesn’t happen because, while MariaDB 1 is restarting, ProxySQL marks this node as shunned. SQL requests are then routed to the other two nodes, and as each one comes back online, the restarts continue—meaning hostgroup 0 is always available. Thanks again!","commit_id":"49cebeaf4dfaf018b06016601aa54c00cbcfa13d"}]}