MySQL takes time to restart, causing successive steps to sometimes fail about ansible-role-mysql

[ergonlogic]

It looks like there is a mysqld process running, so perhaps a second one is started at some point...

[ergonlogic]

Could it be that this isn't aggressive enough. Maybe a straight-up pkill mysqld?

[geerlingguy]

Or a wait to make sure it's completely and totally shut down (it might take a while to shut down the first time, but report back that it's stopped before it's actually stopped). I am growing to dislike Debian/Ubuntu tremendously... between weird symlinked and changed-per-implementation paths and the terrible defaults and installation processes, it's an automation nightmare.

[ergonlogic]

So I guess that'd be accomplished with a 'sleep' parameter in the restart mysql handler?

[geerlingguy]

@ergonlogic - I'd still like to know what, in the environment, might be causing this issue. In my entire fleet of prod servers (mostly CentOS 6/7, but a good number (20+) Ubuntu 14.04 and 12.04), I've never had an issue where an instance of MySQL would get stuck :-/

I'm using the same role on Drupal VM, which uses the geerlingguy/ubuntu1404 box (which doesn't have MySQL pre-installed), and nobody else seems to have run into the error.

[ergonlogic]

The intermittent nature of this behaviour makes it difficult to confirm. But, in the original bug report (GetValkyrie/valkyrie#54 (comment)) the issue arises when we call a Drush command on a site right after restarting the mysql service, as part of flushing handlers.

Note that the sleep parameter to the service module is described as:

If the service is being restarted then sleep this many seconds between the stop and start command. This helps to workaround badly behaving init scripts that exit immediately after signaling a process to stop.

I've added a 5-second sleep in my fork of this role (master...GetValkyrie:master) that I've since included in Valkyrie. It seems somewhat heavy-handed, but I'm not in a position to be able to test this myself. Hopefully, one of the users who is seeing this more consistently can provide feedback.

[ergonlogic]

According to GetValkyrie/valkyrie#54 (comment), it looks like adding a 'sleep' to the service restart handler worked.

[geerlingguy]

@ergonlogic - Ah, that makes sense then. I don't have anything run immediately after the restart, so the restart completes successfully by then.

I'll add a short sleep period to my role (5 seconds seems reasonable), since this could potentially affect others using the same init script, and it's not really a harmful change (just 5 more seconds every time the playbook is run... but an okay tradeoff).

[Prospus-ShantiGola]

Could it be that this isn't aggressive enough. Maybe a straight-up pkill mysqld?

Thanks, it works ....after using this comand just start my mysql service straight away " sudo service mysql start" service start working again.