Re: Exchange dies weekly

I forgot to mention that you can check your queue BEFORE the server locks up to determine if you have this problem. There will still be connections with an open connection with seconds thousands...or tens of thousands. No need to wait for it to lock up just to see if this is your problem. In fact, in some rare cases, if you wait for it to lock up, it is so frozen you can't even view queues anymore. So, unless you *just* restarted (which will obviously flush all connections) you can check this anytime.

-Cliff


"Cliff Galiher" <cgaliher@xxxxxxxxx> wrote in message news:#HYBXqu6JHA.1420@xxxxxxxxxxxxxxxxxxxxxxx
One thing that nobody has mentioned yet is ....have you checked your queues before trying to restart?

There is a known bug *IF* you use SenderID filtering that causes Exchange to see connections in an open state even after the remote server has closed the connection. This results in your queue slowly filling up (with very long connect times) and eventually running out of resources and hanging. You *CAN* recover by killing the appropriate service manually, but ultimately you should apply the appropriate hotfix.

Unfortunately MS's documentation on this hotfix makes it hard to find (the usual keywords you'd think would work just don't appear in the description) and it requires Exchange SP2 be installed, so if you haven't done so then there will be some planning involved.

http://support.microsoft.com/kb/927478

-Cliff



"Billy Gareth" <BillyGareth@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message news:BA746044-0C46-4D06-8E9B-1CE217EFB53A@xxxxxxxxxxxxxxxx
Hi,

We have an SBS 2003 box that is relatively new. The server is an IBM Quad
Core with 8GB RAM and 1.5 TB HDD

Recently the users started complaining about once a week of emails not
working. Upon investigation we found that all aspects of Exchange looked
fine, DB was mounted and all looked good. However when we tried restarting
the Information Store the service would simply fail with the service stuck on
the status of "Starting" The only way to resolve this was to restart the
server.

We have run Exmon and it has returned the following:

Performance Issue found on Disk
The ratio of Reads I/Os to Writes I/Os on drive C: was 0.12. Generally, this
ratio should be less than 0.10; a higher ratio indicates that this disk is
used for something other than writing to the transaction log files. For best
performance, the log drive should be dedicated for transaction log files.

Paged Pool Memory
Paged pool memory on server **** is over the error threshold of 220 MB.
Current value: 231 MB.

Non Paged Pool Memory
The maximum non-paged pool memory on server **** is over the warning
threshold of 100 MB. This may cause system instability. Current value: 107 MB.

There were also some network errors in Exmon but they seemed to be more
related to Trend Micro

While I can obviously read on on the solutions, I would be interested in
some wisdom on why this would happen and best ways to resolve it. Just 'cause
I do it a certain way doesn't mean that it is the best.

Thanks

Billy

.

Relevant Pages

  • Re: Queue qsize = unreliable?
    ... > How I notice a blocking function? ... What I meant was that the fact that a lock is being acquired in the ... > implementation of a queue without any locks. ... support that many connections anyway, and even if you could, using ...
    (comp.lang.python)
  • Re: One Remote Delivery queue is filling up
    ... Check the properties of the queue in question. ... Unable to bind to destination server in DNS ... The receiving server dropped the connection because of a rule. ... >>Outbound connections to 3000 and messages per domain to ...
    (microsoft.public.exchange2000.protocols)
  • Re: How to programmatically avoid deadlock?
    ... pointer or object reference onto/off a queue, ... of lock sequences becomes... ... transaction object is called with the transaction object as a parameter. ... Certainly the lunification-level increases if the number of connections is ...
    (comp.programming.threads)
  • Perplexed by SMTP Current Sessions queue
    ... Checked the current sessions queue and there were over 50 sessions all ... The connections were via our offsite Backup mail server. ...
    (microsoft.public.windows.server.sbs)
  • Re: Users losing "Net use lpt1:". Why?
    ... Is SHARON a server? ... Windows has an allowance for this - it lets you setup persistent connections ... server PRINTQUEUES with queue INVOICES, ...
    (microsoft.public.windowsxp.network_web)
Loading