Mar 20, 2012 2:46 PM
Virtual appliance sucks with 2GB
-
Like (1)
I have the VM appliance - 4.1.70-1455. It is setup with 2GB RAM. It quickly becomes unusable, even with only about 10 devices. I have currently upped the memory to 3.5 GB - see how that works.
BTW - the great thing about Zenoss was always - no java process. Now I have a java process running jetty which is the usual java CPU / memory hog!! This is serious retrograde step!
Cheers,
Jane
Jane, the memory requirements of Zenoss have significantly changed vs. older versions. Our minimum commercial installation requirement is 8 gigabytes of memory. As for the inclusion of Java now, as I understand it, our event processing system has been completely rewritten in the java framework for higher performance - I hope that in your testing you find this to be net beneficial.
Having given the VM 3.5 GB, it now soaks all that (having left it overnight) and is using nearly half-a-Gig of swap. Main culprits are mysqld, runzope and java.
I have 16 devices - fairly standard stuff with fairly standard monitoring.
I have a nasty suspicion that if I give this VM 4GB then it will soak all that and start on swap again. *GB may be OK for Enterprise but many Core customers are monitoring handfuls-to-small-hundreds of devices and 8GB seems unreasonable as a starting point?
Cheers,
Jane
Thanks for the feedback, I completely understand your concerns here, I've raised this and we're working on addressing this issue. Keep an eye out for these changes to make it into the appliances within the next 1-2 weeks.
The Installation manual for Zenoss 4.2.0 states a minimum memory requirement of 4 GB for up to 250 devices.
This seems to be outright wrong since our 4.2 VM repeatably locks up with an out of memory condition after a few hours. We have imported 120 devices on a machine with 6 GB RAM, 2 GB Swap and 2 CPUs.
Java/zenweventserver alone grabs constantly 3.4 GB and zenmodeler uses a whopping 2,4 GB. :-(
The monitored memory consumtion of the zenoss daemons does not increase over time, but what puzzles me is that immediately before the machine becomes unresponsible I can see an increaing number of multiple instances of zenjobs, zenhubworker and WindowsMonor daemons that eat up all the memory. Is this normal?
Just upgraded to 8 GB and waiting what happens...
Hey Folks - It's possible to scale the java heapsize back by doing the following. It shouldnt impact performance to much with a low number of devices.
1. vi /opt/zenoss/bin/zeneventserver. Append an -Xmx512m option to the DEFAULT_ZEP_JVM_ARGS. So that it looks like this:
DEFAULT_ZEP_JVM_ARGS=${DEFAULT_ZEP_JVM_ARGS:-"-server -Xmx512m"}
2. zeneventserver restart as the zenoss user to apply these new settings.
There is also script called mysql tuner that might help with your mysql pain Jane.
Hi Foorhees, thanks for the tip how to limit java memory. It brought down zeneventserver memory consumtion from 3.4 to 2 GB and with a total of 8 GBytes of RAM the machine seems to be at least stable now.
Optimizing Memory Usage for Zenoss 4: http://wiki.zenoss.org/Optimizing_Memory_Usage
Follow Us On Twitter »
|
Latest from the Zenoss Blog » | Community | Products | Services Resources | Customers Partners | About Us | ||
Copyright © 2005-2011 Zenoss, Inc.
|
||||||||