RE: What to monitor on an Exchange 5.5 server

  • From: "Mulnick, Al" <Al.Mulnick@xxxxxxxxxx>
  • To: "'[ExchangeList]'" <exchangelist@xxxxxxxxxxxxx>
  • Date: Thu, 20 May 2004 09:08:57 -0400

For finding the pulse of how my 5.x environment is running, I like to break
it up into the four food groups: CPU, DISK(I/O), Network, and Memory.
Usually in that order of importance.  Here's why.

I'm interested to know that my CPU doesn't go above 80% and stay there for
many consecutive sample intervals.  If it does, I may have a problem with
either the server or my monitoring sample interval ;) 
I'm interested to know what my Disk I/O is especially on my log file drive
and my database drives.  If I've separated I/O, I likely have my db's on
separate drives (priv on one set of spindles RAID 5 or 1+0 for fast read),
Dir on a separate set of spindles if large enough to warrant it, and pub on
it's own set of spindles if not on it's own machine (preferred in 5.x).  I
would also like to see my MTA separated if large enough implementation. I'm
most concerned that my log file drive is not constrained, then my DB
(they're related), then my MTA, etc.  I also want my swap file to be happy.
So I monitor drive write/read queue and make sure it falls within acceptable
levels (I'll mention where you can find that information later).
Next, I care about my network.  Do I have a lot of errors?  Do I have a
large queue?  In short, is it a bottleneck?

Finally, I care about memory.  Memory is something that Exchange will try to
use all of that it can find. It's designed that way.  Of course, because of
the OS and other limitations, I don't want to go over about a 1gb phys
memory config.  It can degrade perf if I go over that on a 5.x/NT box. 

Of course I care about the other health as well.  I want to see the queue
levels, the error logs, disk space remaining, and other operational level

As for thresholds, most of them are related to the OS.  Check out the NT 4
reskit for information on what the levels are supposed to be for a healthy
server for the four "food groups".  For the Exchange thresholds
specifically, watch the queues for any hang ups that shouldn't be there.

Finally, it would be a lot easier, if you're a large shop to get a
monitoring application that understands what the recommended thresholds are
out of the box.  That's what NetIQ does for you (AppManager), Spotlight, and
MOM (with the 5.x module). I'm sure there are others, but those come to mind
off the bat.


-----Original Message-----
From: Susan Tackett [mailto:susan_tackett01@xxxxxxxxxxx] 
Sent: Thursday, May 20, 2004 8:49 AM
To: [ExchangeList]
Subject: [exchangelist] What to monitor on an Exchange 5.5 server

I have an Exchange 5.5 server running on an NT4 SP6a OS and want to set up
some best practices for monitoring it.

I have been looking at packet monitoring and dramatic increase in container
size, but I need to setup guidelines/procedures for what should be monitored
in terms of CPU utilization, RAM utilization, and Exchange-specific
variables? What are some are the most important red flags to look for?

I have done several searches on, but have been unable to find
the reference material that I need.

Thank you for your help-


Is your PC infected? Get a FREE online computer virus scan from McAfee(r)

List Archives:
Exchange Newsletters:
Exchange FAQ:
Other Internet Software Marketing Sites:
Leading Network Software Directory:
No.1 ISA Server Resource Site:
Windows Security Resource Site:
Network Security Library:
Windows 2000/NT Fax Solutions:

Other related posts: