Page MenuHomePhabricator

s52584 is taking over half of the available connections on toolsdb
Closed, ResolvedPublic


Persistent connection model is not allowed on labsdb hosts for the reasons expressed on the Connection handling policy.

Your connections are currently taking half of the used connections, impacting other users:

root@labsdb1005[(none)]> pager grep -c s52584
PAGER set to 'grep -c s52584'
root@labsdb1005[(none)]> SHOW PROCESSLIST;
164 rows in set (0.00 sec)

Please disable connection pooling on your tools, or request a dedicated mysql instance to use in any way you like (toolsdb is a *highly* shared resource, so limits have to be enforced). Enforcing it myself (for example, limiting your max connections to a handful of them) may cause you an outage, so I would prefer if it was you that solved this issue yourself. Thank you for your understanding.

Event Timeline

jcrespo created this task.Apr 11 2017, 12:12 PM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptApr 11 2017, 12:12 PM
jcrespo updated the task description. (Show Details)Apr 11 2017, 12:54 PM

Hi, @jcrespo. Per your comment at T138283#2412013, I disabled connection pooling only for the replicas and -somehow- convinced myself that the tools-db config was good to keep as it was. Excuse my (severe) lack of understanding, I've just killed all connections and corrected the settings for any db resources that were making use of pools.

Could I suggest rephrasing the Given the number of potential users for the database replicas [...] part in the connection handling policy so that it's made totally clear that previous statements do, in fact, apply to tools-db, too?

jcrespo closed this task as Resolved.Apr 11 2017, 1:29 PM
jcrespo assigned this task to PeterBowman.

Excuse my (severe) lack of understanding

No, clearly it was my fault for not being 100% clear. Also, the toolsdb server has increased its usage in the last months, so it has become more of a deal lately. It will also be helpful to avoid stale connections in case the server is failed over, which happened some weeks ago. I will add your suggestion to the wiki, thanks.

Thank you really for your quick response!