opendevreview | Lingxian Kong proposed openstack/trove master: Fix resize instance of replication cluster https://review.opendev.org/c/openstack/trove/+/800019 | 01:04 |
---|---|---|
opendevreview | Lingxian Kong proposed openstack/trove master: Restore configuration after resizing instance https://review.opendev.org/c/openstack/trove/+/800019 | 01:09 |
__ministry | lxkong: I was view new your patch about fix error when resize, but i think if we do that, when i do resize, it will take long time. What problem which we met if we do as my path? | 02:10 |
lxkong | __ministry: ah, I see your point, let me do some test | 04:56 |
lxkong | __ministry: because lots of functions are going to be affected by that change | 04:57 |
lxkong | which I am trying to avoid | 04:57 |
__ministry | when that function need override old config? I can't find :( | 04:58 |
__ministry | what are contexts in here? | 04:59 |
lxkong | __ministry: look for the places that call save_configuration() | 05:02 |
lxkong | and `reset_configuration()` | 05:03 |
lxkong | the patch is going to change the behavior of all of those places | 05:03 |
lxkong | the original purpose of `save_configuration` is to reset configuration | 05:04 |
lxkong | which needs to remove all current user defined configs | 05:04 |
lxkong | now you changed that, that may help resize, but may affect others | 05:04 |
opendevreview | Lingxian Kong proposed openstack/trove master: Release note for adding iptables-persistent package https://review.opendev.org/c/openstack/trove/+/800158 | 05:26 |
__ministry | I saw, when I save configuration, if old config still in here, and we do update configuration group, it still effect to old file =)). | 07:57 |
__ministry | Let me test your patch. | 07:58 |
opendevreview | Lingxian Kong proposed openstack/python-troveclient master: Fix detach configuration from instance https://review.opendev.org/c/openstack/python-troveclient/+/800166 | 08:48 |
opendevreview | Merged openstack/python-troveclient master: Fix detach configuration from instance https://review.opendev.org/c/openstack/python-troveclient/+/800166 | 10:31 |
lxkong | __ministry: I'm also verying yours | 10:45 |
lxkong | s/verying/verifying | 10:45 |
lxkong | __ministry: what do you mean by "we do update configuration group, it still effect to old file"? | 11:15 |
lxkong | How did you test that? | 11:16 |
__ministry | Oh, I using my patch to create an instance and resize it | 11:17 |
lxkong | have you tested resizing the replica using your patch? | 11:17 |
__ministry | after i do update configuration for instance, and I was see config value was changed. | 11:17 |
__ministry | lxkong: yep, I was tested. | 11:18 |
lxkong | so expected? | 11:18 |
__ministry | my expected that are: when I resize a instance, I think about deep dive, it just is rebuild flavor and keep root disk. | 11:21 |
__ministry | but we doing that are change data of root disk. | 11:21 |
__ministry | so, my expected about resize instance take less time. because, if we reboot database engine, it is maybe to take long time to start | 11:23 |
__ministry | it take resize flow long time too. | 11:23 |
lxkong | restart database is just a docker container restart, which is actually fast. The point is not restarting, I'm looking at why the user defined configs are removed in save_configuration. Maybe I will do some changes based on yours. | 11:26 |
lxkong | so basically, we need to verify: | 11:26 |
lxkong | 1. create a replication cluster with at least 2 instances. | 11:26 |
lxkong | 2. resize primary, see if that will break the replication function | 11:27 |
lxkong | 3. resize replica, check if it is still working as a replica | 11:27 |
lxkong | 4. resize a stand-alone instance (with configuration attached), see if the configuration still exists | 11:27 |
lxkong | 5 change the configuration of the stand-alone instance, see if the configuration is correct inside the instance. | 11:28 |
__ministry | yep. | 11:29 |
__ministry | thank you. | 11:30 |
opendevreview | Merged openstack/trove master: Add package iptables-persistent inside guest instance https://review.opendev.org/c/openstack/trove/+/800130 | 13:40 |
opendevreview | Merged openstack/trove master: Release note for adding iptables-persistent package https://review.opendev.org/c/openstack/trove/+/800158 | 21:04 |
opendevreview | Lingxian Kong proposed openstack/trove stable/wallaby: Add package iptables-persistent inside guest instance https://review.opendev.org/c/openstack/trove/+/800182 | 21:10 |
opendevreview | Lingxian Kong proposed openstack/trove stable/victoria: Add package iptables-persistent inside guest instance https://review.opendev.org/c/openstack/trove/+/800311 | 22:11 |
lxkong | __ministry: with your patch, I met with the error inside the guest intance `2021-07-09T21:36:05.220833Z 0 [ERROR] Fatal error: Can't open and lock privilege tables: Table './mysql/user' is marked as crashed and should be repaired` when resizing replicas. | 22:20 |
lxkong | I have reproduced many times. | 22:20 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!