[pgpool-general: 1210] Re: Semphore errors after adding/removing hosts?

[Nick Webb]

On Fri, Nov 30, 2012 at 5:18 PM, Lonni J Friedman <netllama at gmail.com> wrote:
> I've seen this behavior as well (after dettaching & reattaching nodes)
> with 3.2.1.  However, when I experienced it, it was after dettaching
> all the nodes at the same time, which I was later told was not
> supported without restarting the master pgpool daemon.

I only detached one node at a time.  There are two nodes, and I
verified that the status of the newly attached node was != 3 before
detaching the second node.

By the way, I'm using version 3.1.5... I started with 3.1.3, then
upgraded to .5 to see if that would help (it didn't).  I was hesitant
to upgrade to 3.2.x rashly, but perhaps I can do that this weekend
given a bit more time when the system is slow.

>
> How many nodes do you total ,and how many were active at any given time?

2 nodes, 1 was always active.

>
> On Fri, Nov 30, 2012 at 4:41 PM, Nick Webb <nickw at redwireservices.com> wrote:
>> Hello,
>>
>> I'm fairly new to pgpool, but am already quite a fan, thanks for the great
>> software!
>>
>> When using the pcp_detach_node command followed by pcp_attach_node after
>> doing some maintenance on a backend, I start getting errors like this in the
>> log:
>>
>> Nov 27 03:09:02 PGLB pgpool-X[32282]: semop(id=884737) failed: Invalid
>> argument
>> Nov 27 03:09:02 PGLB pgpool-X[10230]: semop(id=884737) failed: Invalid
>> argument
>> Nov 27 03:09:02 PGLB pgpool-X[10230]: semop(id=884737) failed: Invalid
>> argument
>>
>> These continue over and over until the process is restarted.
>>
>> I ignored them when I first saw it, as it seemed to function just fine, but
>> a few days later our application could not connect to PGPool very well, it
>> was very slow and most connections timed out despite the system load and CPU
>> being very low.  Connections from the PGPool host to the back-ends directly
>> were also very quick (via psql command).
>>
>> Stopping and starting PGPool immediately returned the service to normal
>> performance levels.
>>
>> Are these two issues linked?  How can I stop from getting the semop error
>> when attaching hosts?
>>
>> It also seems that about every 3rd or 4th pcp_ command I run hangs,
>> seemingly waiting until open connections finish... and I think it finish
>> after a few minutes.  This may be related, as I seem to recall one of these
>> "hangs" when adding the backend back to the pool, then seeing the semop
>> errors (could have been the detach that hung, not sure).