Re: Monitoring IMAP thread count

[Brian Elliott Finley <finley@anl.gov>]

Correction -- I was too hasty in hitting reply.  The "grep -w 143" 
example, will also match an IP address with 143 as on of the components.

Also, if you have the ipv6 stack loaded, it may affect your results too. 
 In our case, it's loaded, but we don't have clients connecting to it. 
 However, this is what we see in our output.  The first line is ipv4, 
the second is ipv6, but they are for the same connection:

tcp        0      0 130.202.101.12:50659    130.202.101.12:7143 
ESTABLISHED
tcp6     0 0 ::ffff:130.202.101.12:7143 ::ffff:130.202.101.12:50659 
ESTABLISHED


So, here's a more refined example, and assumes only ipv4 connections are 
being used for connectivity.  It eleminates the possibility of the port 
number matching an IP address:

  netstat -an | egrep '^tcp .*:7143 +ESTABLISHED$' | wc -l

Cheers, -Brian


Brian Elliott Finley wrote:
> Thanks for the query line.  If you change the "grep" to "grep -w", you 
> can get an accurate count for either 143 or 7143.
>
> On a system with nginx, for example:
>
> % netstat -an | grep -w 143 | grep ESTABLISHED | wc -l
> 1
>
> % netstat -an | grep -w 7143 | grep ESTABLISHED | wc -l
> 1302
>
>
> Cheers, -Brian
>
>
> Steve Hillman wrote:
> > While you can query the Zimbra Admin REST API and get what it *thinks* 
> > are the number of connections, this, in our experience, tends to be 
> > wrong. We've been graphing that number for some time and never saw it 
> > get near the limit, yet at the start of this semester, we started 
> > getting IMAP problems too.  Sure enough, when we poked around, we 
> > found we were typically 100-200 connections higher than what was being 
> > reported.
> >
> > I think the simplest check is this, on each mailbox server: "netstat 
> > -an | grep <IMAPPORT#> | grep ESTABLISHED | wc -l" (that's assuming 
> > you have nginx in front and your IMAP port # on the mailbox servers is 
> > a high port, like 7143. If it's port 143, you'll get false matches)
> >
> > The connections reported by zmstat-chart are also accurate, but 
> > they're not typically designed to be fetched real-time
> >
> > ----- "mem@isc.upenn.edu via zimbra-hied-admins-request" 
> > <mem@isc.upenn.edu> wrote:
> >
> > > Hi folks,
> > >
> > > Lately we've been running into trouble with our IMAP thread counts 
> > > reaching our defined limits, which seemed to be high enough until very
> > >
> > > recently. We'd like to monitor the thread counts and have an alert
> > > when we are getting close to our limit.
> > >
> > > Does anyone have any experience with this? and if so, what are you
> > > doing to monitor thread counts?
> > >
> > > Thanks,
> > > --
> > > Matt Miller
> > > ISC/Network Services          phone:  215-573-3829
> > > University of Pennsylvania   jabber: mem@upenn.edu
> > >                         ----
> > > DA49 9257 E1B9 4C6F C94B  7E3C A2A4 32E3 55F8 E421

--
Brian Elliott Finley
Deputy Manager, Unix, Storage, and Operations
Argonne National Laboratory, Computing and Information Systems
Office: 630.252.4742  Mobile: 630.447.9108