Hello,

I have a cluster of openldap servers behind a load balancer. Openldap
writes its logs to syslog(). Rsyslog grabs them via imuxsock and writes
them to files in /var/log and also ships them to a remote server.
Nothing fancy...

The problem I have is that from time to time, the load balancer ejects a
node from the cluster for a few seconds, which means openldap is blocked
or is refusing new connections. This happens only when there's a
diskfull on the node's /var/log, which is why I'm asking here.

As a workaround, I recently changed my config to have 2 separate action
queues, configured to drop logs when the queue is full
($ActionQueueDiscardSeverity 0 among other settings).

Now using ps/top, I see the 2 "rs:action X" threads, which are sleeping
most of the time. But what strikes me and worries me is that the
"rs:main Q:Reg" thread is in D (uninterruptible sleep) state most of the
time, waiting on "jbd2_log_wait_commit" (which is related to ext4 if I'm
not mistaken).

So I'm wondering:

 - why does the main-queue thread do any disk-related activity at all ?
   Shouldn't it just pass the logs down to the action-queues, and if any
   disk operations have to take place, they would happen in the
   action-queue threads ?

 - how can I be sure openldap will not block because rsyslog is having
   trouble ? It *seems* ok now with the capped action queues, but this
   thread in D state makes me wonder.

NB: for the sake of understanding, I have disabled disk-assistance on
all the queues at the moment. And iostat indeed shows almost no activity
on the volume holding rsyslog's $WorkDirectory.

This is running 5.8.10 (version shipped with RHEL6). I know it's old,
and I'm open to upgrade it if needed.

Thanks !

Marc
_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com/professional-services/
What's up with rsyslog? Follow https://twitter.com/rgerhards
NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of 
sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE 
THAT.

Reply via email to