General discussion

Locked

Best practices to monitor Exchange 5.5

By jasonhiner Moderator ·
This question is being posted by the TechRepublic staff. The response(s) from this post will be turned into an article on the TechRepublic site. This is a winner-take-all scenario and the winner will be the person who provides the best and/or most complete answer in our judgment. If there are multiple correct answers we will select the one that was posted first. In addition to offering a generous amount of TechPoints, the winner will also receive a free piece of TechRepublic gear (e.g. a coffee mug, a desk flag, or another item sporting the TechRepublic logo). This competition ends at 11:59 PM on Sunday, May 23. Good luck and let the best techie win!

HERE IS THE SCENARIO: A network administrator has inherited an Exchange 5.5 server and wants to set up some best practices for monitoring it. What does the admin need to monitor in terms of CPU utilization, RAM utilization, and Exchange-specific variables? What are some are the most important red flags to look for?

This conversation is currently closed to new comments.

23 total posts (Page 2 of 3)   Prev   01 | 02 | 03   Next
Thread display: Collapse - | Expand +

All Comments

Collapse -

by csinath In reply to Best practices to monitor ...

Hmmmmmmm....I'll make a try. I think you should do an upgrade.

Collapse -

by jasonhiner Moderator In reply to

Poster rated this answer.

Collapse -

by zarina In reply to Best practices to monitor ...

From personal experience,
1.Put your Exchange 5.5(database, program etc.,) on a separate location - partition, hard disk from any other softwares or data - if you can afford it-a separate server. Have a scheduled offline backup(an exact copy) and don't depend on your 3rd party backup software. Exchange 5.5 advantage is this-you can restore it by just duplicating the exact structure.
2.Keep the database size manageable. All the recovery procedures will work perfect on considerable db size. Keep ample free space on the hard disk/partition (twice or more than the db size). Those rec. proc won't work too with constraint space.
3.Monitor your event log for MTA alerts-especially.You can set according to all MS Exch5.5 defaults.
4.Look out for CPU% usage as there is a known bug with Exc.5.5 that halts operation and bog down with 100%CPU usage.
5.Get a copy of O'Reilly 'Managing Microsoft Exhange Server' by Paul Robichaux-1999 at hand. This book has a wealth of information on Exchange 5.5.
Good Luck.

Collapse -

by jasonhiner Moderator In reply to

Poster rated this answer.

Collapse -

by jburton In reply to Best practices to monitor ...

Monitor hardware ?use SNMP to alert on server, hard drive, power supply, or UPS failures. Check the system event logs daily for hardware errors.
Monitor back ups -use application event logs (and your back up logs if available) to check for successful back ups. Make sure the transaction logs are deleted after a checkpoint. Otherwise, the log files will eventually fill up your log disk since circular logging should be turned off.
Monitor disk space. Exchange shuts down in self defense if it runs out of disk space.(-510 error.)
Monitor application event logs for ?online database defrag successful? informational events.
Monitor for -1018 and/or -1022 errors in the application event log, that's a red flag showing physical database corruption. If you see these, resolve your hardware problem and restore the database.
Set up an independent ?Mail Ping? to check sending and receiving from Exchange.
Use performance monitor(NT4) or system monitor(W2K):
- Processor -%Processor Time: Consistently over 75% (each processor)-red flag.
- Process - %Processor time: Use this to identify the process causing high CPU use.
- Memory - Pages/sec: Consistenly averaging greater than 10 -red flag.
- Memory - Commited bytes: Must not exceed the overall size of the pagefile.
- Process ? Page faults/sec: Use this to identify the process using up the memory.
- MSExchangeMTA ? Work Queue Length: if this counter is greater than zero for a sustained time frame, -red flag.
- MSExchange Private & Public ? Average Time for Local Delivery: Should never be above zero for longer than a few seconds.
- MSExchangeDS ? Pending Replication Synchronization ? more than one Exchange server, replication requests should decrement to zero. If not ?red flag.
- IMS monitor queues to make sure mail is flowing.
Use the pre-defined performance charts installed by Exchange (Server Health, Server Load, Server Users, Server Queues and IMS usage) to create a baseline of normal activity and compare weekly.

Collapse -

by jburton In reply to

-And monitor the Exchange services.

Collapse -

by jasonhiner Moderator In reply to

Good information.

Collapse -

by mjost In reply to Best practices to monitor ...

The first thing to do is sit down and develop simple strategies for service, usage, configuration and problem management. Service and usage management may be less important but special attention should be paid to configuration management. This is the key to avoiding problems in the first place. Simple good management practices like maintaining server installation and upgrade documents are the cornerstone of configuration management. For service management simply make sure that a clear and concise document is created that specifies what the system will deliver to the user. For example this should specify service features such as maximum mailbox sizes and maximum message sizes.

Measure performance and usage. Performance degradation has two possible causes - system breakdown or increased load. Next create some key Exchange management and performance monitoring statistics.

Use Server analysis to:

? Creating a baseline of current use.
? Monitoring use over a period of time.
? Analyzing data to find and resolve abnormalities in the system use.
? Determining expected response times for specific numbers of users and system use.
? Determining how the system should be used.
? Determining when to upgrade the system, or when to add additional system resources.

Here is some basic information on some of the counters that that were asked to be monitored as part of the data collection and analysis process.

Collapse -

by mjost In reply to

System CPU(s)
Abnormally high CPU utilization is normally the side effect of a problem with a separate system component such as the disk or network subsystems. Continued high CPU utilization i.e. 80%+, has also been seen to be caused by software issues. Note: When overall processor usage is showing a consistently high value it should be compared against how many users are currently connected to the system. Simply add the "MSExchangeIS ? User Count" counter to correlate this information on the performance monitor chart in question.

System Memory
Discussed below are counters that will assist in detecting whether memory is a bottleneck on an Exchange system. If Exchange performance is impacted in any noticeable way then the first area that needs to be monitored is the system memory. CPU and disk I/O time could be showing up as a bottleneck as a result of them trying to cover up a system memory deficiency. The key to monitoring and determining memory performance is to determine how much a system is paging data in and out of memory.

Exchange Specific items to monitor.

The performance monitor also can be configured to monitor the key components of Exchange. Configure server and link monitors to identify failed services and message delivery delays. This can all be implemented on the server console itself i.e. log onto the server, configure the performance monitor alert view appropriately and start the server and link monitors, then lock the console so that the monitoring in continuous.

Collapse -

by mjost In reply to

A number of pre-defined performance monitor charts are included with the standard installation of Exchange Server. The charts installed by Exchange setup are:
?Server Health: This chart is updated every second. As its name implies it monitors the overall health of an Exchange server. It does this by charting Windows NT counters for total percentage CPU time and the percentage CPU time of each Exchange core process. This chart is good for immediately identifying a process that is utilizing too much CPU time.
?Server History: This chart gives a general overview of system performance. It charts the number of users currently on the system and it also charts the amount of outstanding messages still waiting to be processed in the internal queues.
?Server Load: This detailed chart goes deeper into the Exchange server and tracks items such as address book usage and the number of messages being submitted and delivered among others.
?Server Users: This chart simply indicates how many users currently have a connection to the Exchange server.
?Server Queues: This chart tracks the internal queues on the Exchange server. It charts all messages that are stored in these queues including both the Information Store and MTA work queues.

For continuous monitoring purposes, only few of the counters need to be used. These critical "Exchange Server Health", counters are as follows:
Object/Counter
Memory/ Pages/sec
LogicalDisk(s)/ Free Megabytes
Processor/ %Processor Time
Redirector/ Bytes Total/sec
MSExchangeMTA/ Work Queue Length
MSExchangeIS/ User Count
MSExchangeIS Private/ Messages Submitted/min
MSExchangeDS/ Pending Replication Synchronization
These few counters will identify the majority of server performance problems. In addition to these there are 3 charts installed for the Internet Mail Service (IMS) that are used to track queue and traffic information for the IMS. Use a wider range of counters to help pinpoint the problem further.

Back to Windows Forum
23 total posts (Page 2 of 3)   Prev   01 | 02 | 03   Next

Related Discussions

Related Forums