Re: Tracing TCP connection numbers on POA

Michael Bell

unread,

Feb 4, 2010, 11:03:27 AM2/4/10

to

On 2/4/2010 2:56 AM, Arnold71 wrote:
>
> Hi all,
>
> I'm troubleshooting a performance/utilization issue on a GW 7.0.3 POA
> running on Netware 6.5.
>
> On some days (always tuesday and thursday) the server has a constant
> utilization of about 90-95%. When this occurs, I switch the POA logging
> to Diagnostic and the screen is filled with "Dispatching message to
> connection #18 (TCP)". The next day, the connection number is different
> obviously.
>
> I think there might be a user out there who is running something other
> than the GW client to talk directly to the POA but I can't trace the
> exact user. Is there a way to track the connection number to a real
> user? I figured that it isn't related to the connection number in the
> Netware monitor, is it?
>
> Arnold...
>
>
No it's not. But I think there is a list of connections on the http
monitor...

Dave Parkes

unread,

Feb 5, 2010, 6:01:20 AM2/5/10

to

Maybe that is the connection to the MTA ?. With Diagnostic logging, you do
get buried with log entries, verbose is probably better. But use the http
console to see if any gwchecks are running or maybe even just indexing

Cheers Dave

--
Dave Parkes [NSCS]
Occasionally resident at http://support-forums.novell.com/

Dave Parkes

unread,

Feb 5, 2010, 7:37:34 AM2/5/10

to

Next time the POA is busy, check the HTTP console to see if that shows any
threads there as busy. It could be something as simple as someone doing a
lot of purging on their account, also deleting an account is an intensive
process.

Traffic to the MTA is unlikely to be workstation/account related as that
is only either admin messages or out of PO message transfer, But checing
that HTTP console should give you a clue, you can check the MTP status
from one of the tabs there.

Also, the POA will back off processing if the server is really busy doing
something else. If the POA is busy, while the server itself is not, that
could mean that you are hitting a maximum number of threads somewhere
within the POA config while you actually do have spare capacity to
increase that number of threads, again the HTTP monitor should show you
that info.

Dave Parkes

unread,

Feb 9, 2010, 6:51:15 AM2/9/10

to

Maybe something to do with how the GW files are being accessed. Perhaps
the delay is there somewhere. What shows as busy in the NRM monitors for
the servers themselves ?

Dave Parkes

unread,

Feb 9, 2010, 12:38:11 PM2/9/10

to

Out of interest, was that just an IMAP connection or something cleverer
through GMS or similar

Dave Parkes

unread,

Feb 10, 2010, 7:23:46 AM2/10/10

to

I'd think it was IMAP then. TID 7004663 might be slightly helpful, move
the IMAP connections to the GWIA and turn IMAP off at the post office.

Dave Parkes

unread,

Feb 11, 2010, 11:05:29 AM2/11/10

to

I can't remember what protocols Intellisync can use, but if you are
already using GMS, then that might be an alternative

Dave Parkes

unread,

Feb 12, 2010, 6:17:25 AM2/12/10

to

To Arnold71

>But for now, I'm going to enjoy my long weekend and I'm going to stop
>worrying about all of this! ;-)

Seems eminently fair <g>