While discussing the InnoDB crashes with MariaDB devs, they were asking whether we were running periodically check table.
Maybe we should start running a check table on a weekly (or whatever) basis and report on those tables that are corrupted, at least on backup source hosts.
This would allow us to catch possible index or any other corruptions beforehand.
I recloned db2125 a few days ago, and while running it across all its wiki, I have noticed some indexes corruptions on some tables.
They don't necessarily need to be present on the backup source hosts (I haven't checked), they could be the result of the upgrade, crashes or any other thing, but ensuring we know our snapshot status would be good.
First manual run:
- db1095:s2
- db1095:s3
- db1102:x1
- db1116:s7
- db1116:s8
- db1139:s1
- db1139:s6
- db1140:s1
- db1140:s6
- db1145:s4
- db1145:s5
- db1150:s4
- db1150:s5
- db2098:s2
- db2098:s3
- db2097:s1
- db2097:s6
- db2099:s4 Index 'globalimagelinks_wiki' corrupt
- db2099:s5
- db2100:s7
- db2100:s8
- db2101:x1
- db2139:s4
- db2139:s5
- db2141:s1
- db2141:s6