Closed weiss closed 1 month ago
Been seen these for years, on PG: https://github.com/processone/ejabberd/issues/3698 what's different now?
Been seen these for years, on PG: #3698 what's different now?
Looks different to me, yours are query timeouts, and I guess you don't get rid of them by disabling prepared statements?
In my case, there's no actual error message (<<>>
).
you don't get rid of them by disabling prepared statements?
this was not suggested no
In my case, there's no actual error message (<<>>).
indeed, looks odd
@weiss Could you see if updating p1_mysql helps here? There was couple fixes in prepared statements after last release
@prefiks .131 means they are missing just ~10 commits, aren't those fixes included?
No, we use tagget/hex released version of p1_mysql in ejabberd git, but there were couple changes after that in p1_mysql repo.
Could you see if updating p1_mysql helps here? There was couple fixes in prepared statements after last release
Looks good so far, no errors after about 30 minutes, I'll reopen if I do run into problems again. Thank you!
Environment
SQL configuration
Errors from error.log
Bug description
I'm running into failures storing MAM messages when prepared statements are enabled with MySQL/MariaDB (which is the default). Unfortunately I don't know how to reproduce. My impression is that it's unrelated to the MAM payload: I checked some of the failed cases, the payloads don't look unusual in any way, and I can't see an obvious pattern. However, it happens once every few minutes on our production server, so I could easily test (debugging) patches.
I'm assuming it's not actually specific to MAM (it's probably just MAM being a busy table), as I've seen one or two cases of PubSub storage failures as well, but I'm not entirely sure here.
Setting
sql_prepared_statements: false
makes the errors go away.