[support] Druapal server fatal error: "MySQL server has gone away", but DB server is still up & accessible. Help troubleshooting?

[grantksupport at operamail.com]

Hi

On Sat, Jun 14, 2014, at 06:33 AM, Jamie Holly wrote:
> When it happens, run:
> 
> mysqladmin version
> 
> And check the uptime. Even though you are seeing the tables with 
> mysqlshow, that doesn't mean that MySQL didn't die and then restart 
> automatically when the error hit.

It does, apparently:

mysqladmin version
	mysqladmin  Ver 9.1 Distrib 10.0.11-MariaDB, for Linux on x86_64
	Copyright (c) 2000, 2014, Oracle, SkySQL Ab and others.

	Server version          10.0.11-MariaDB-log
	Protocol version        10
	Connection              Localhost via UNIX socket
	UNIX socket             /var/cache/mariadb/mariadb.sock
	Uptime:                 12 hours 12 min 40 sec

	Threads: 1  Questions: 2  Slow queries: 0  Opens: 0  Flush
	tables: 1  Open tables: 63  Queries per second avg: 0.000

drush -v site-install standard -y ...

mysqladmin version
	mysqladmin: connect to server at 'localhost' failed
	error: 'Can't connect to local MySQL server through socket
	'/var/cache/mariadb/mariadb.sock' (111 "Connection refused")'
	Check that mysqld is running and that the socket:
	'/var/cache/mariadb/mariadb.sock' exists!

mysqladmin version
	mysqladmin  Ver 9.1 Distrib 10.0.11-MariaDB, for Linux on x86_64
	Copyright (c) 2000, 2014, Oracle, SkySQL Ab and others.

	Server version          10.0.11-MariaDB-log
	Protocol version        10
	Connection              Localhost via UNIX socket
	UNIX socket             /var/cache/mariadb/mariadb.sock
	Uptime:                 5 sec

	Threads: 1  Questions: 1  Slow queries: 0  Opens: 0  Flush
	tables: 1  Open tables: 63  Queries per second avg: 0.200

> You can try start mysql with -log-warnings=2, which has MySQL log more 
> info on problems, especially connection issues.
> 
> If none of that helps, then start mysql with the --log={file name} 
> option. This will log all queries to the file you specify. You can then 
> see which query it is dying on.

also did that

suspecting that this _was_, as you've pointed out, the DB svr, I posted
the query detail here:

  https://lists.launchpad.net/maria-discuss/msg01703.html

I'm not yet clear whether 'they' will help deal with this as it does
involve Drupal, and I'm not yet able to narrow it down.

> One other thing that you can try is changing the DB host from localhost 
> to 127.0.0.1. Sometimes you can have weird problems on the server that 
> causes localhost to have problems and a common reason for "mysql has 
> gone away" is hostname problems (log-warnings will help identify this).

same error(s) using '127.0.0.1' vs 'localhost'

> Unfortunately this is one of those problems that is a PITA to track 
> down. It's not Drupal or even PHP, but rather MySQL. I recently had to 
> track the same problem down on a Java application and it added to the 
> gray hairs on my head.

Yep.  I have just a suspicion that it involves more than one ... We'll
see.

Thanks.