Windows & Linux System Performance Monitor
I was posed these questions via my customer and was wondering if I could get some clarity into the matter as I have never worked with these monitors before.
- How well these monitors scale. I.e., can we expect that they would work well with a monitor configured to run against 1000 or more servers? I presume it is multi-threaded / asynchronous and that it kicks off the monitoring request to all servers are approximately the same time, based on the schedule.
- Related to the question above, can we obtain details regarding where large deployments have been made previously by other customers and how many servers they monitor?
- Is there a way that we can be notified (alert or dashboard) of situations in which a metric is not obtained from a monitored server for a specific time frame… i.e., report on all servers that have not had successful metric retrieval for > 1 hour.
Cheers,
Sanj
Best Answer
Hello Sanj,
Based on personal experience on our client's site, let me address your points above in a dynatrace 5.6 environment:
- These monitors run from collectors which means that you will be limited to the collector performance. We have over 1500 servers (mainly linux) being monitored every minute with performance, service status, or process monitors. This puts some stress to the collector, so we have 10 collectors dedicated to monitors (no agents here, this is not recommended). I have read/heard/being told that you should not put more than 300 monitors per collector. We have seen between 150 to 200 monitors per collector, but all depends also on the schedule you will use, for us anything over 200 shows that collectors start suffering some performance degradation. In your case, I would suggest AT LEAST 4 collectors for 1000 servers.
- I can only speak for my customer.
- Unfortunately I don't think dynatrace have this capability.
Hope this helps you a little bit
Regards,
Santiago
Hey Santiago,
Thank you for the reply, greatly appreciated. I will look for the documentation to see how many monitors can run per collectors but this was exactly the answer I was looking for.
Cheers,
Sanj
Hi there
Depending on what you want to monitor, would it not maybe make sense to roll-out the dynaTrace Host monitoring Agent? I know there is a cost involved. We monitor about 4000 servers in various countries and operating systems. We have a lot of problems trying to use the monitor in our environment, with the biggest problem being firewalls, login access to servers and service accounts etc.
Thanks
Hey Anton,
This would be the ideal situation but my customer is first investigating this method to see if its viable(this is a POC). We have a couple of geographic locations so your concern might prove to be extremely valid in my situation.
Thanks,
Sanj
In addition this is currently the documentation about the Windows Performance Monitoring regarding version:
Most other Measures are calculated by dynaTrace. Those retrieved by the Windows Performance Monitor are not retrieved by an dynaTrace Agent. They are directly calculated by the Monitor. Performance Counters is available on most versions of Windows starting with Windows 95 and Windows NT 4.0. Different Measures may be available for different operating systems.
as per Windows Performance Monitor#Measures
Is there an official list for the windows versions. The "Most versions of Windows" is not thorough enough for my customer.
Cheers,
Sanj
