[SAC] PostgreSQL on TracSVN machine over max connections

Committees SAC mailing list
1

The PostgreSQL server on TracSVN VM reports having reached
max connections (200, per current configuration).

It's been logging this for days, after just one day from start:

  2018-05-22 09:06:03 GMT FATAL: the database system is starting up
  [...]
  2018-05-22 09:06:03 GMT LOG: database system is ready to accept connections
  [...]
  2018-06-26 08:22:01 GMT FATAL: sorry, too many clients already

Logs do not contain any information about clients.
According to netstat there are only 2 TCP connections (which I think
we should forbid, unless there's a reason not to) so I guess all
others are done via unix sockets.

This is PostgreSQL-10 running, dunno since when.
Can anyone join #osgeo-sac to help with fixing the issue ?
For now I'll restart it

--strk;

2

On Tue, Jun 26, 2018 at 10:30:09AM +0200, Sandro Santilli wrote:

According to netstat there are only 2 TCP connections (which I think
we should forbid, unless there's a reason not to) so I guess all
others are done via unix sockets.

I found that at least one of the 2 TCP connections was coming
from Gitea, which is now fixed to use unix socket:
https://git.osgeo.org/gitea/sac/gitea-config/commit/a5db6075
(you need authorization to see)

All the other connections were from Trac,
for which I filed a ticket about:
https://trac.osgeo.org/osgeo/ticket/2176

During the audit I found we're also using a very unsecure
configuration of PostgreSQL which should be fixed ASAP.

Martin, is your contract still ongoing ? Can you look at it ?
Fixing Trac access would be the first step:
https://trac.osgeo.org/osgeo/ticket/2175

--strk;