I added a provisioning group with 170 nodes and the services icmp and wmi to my setup. That worked fine over one hour. Data got collected - everything fine. After one hour opennms needed more and more filehandles and after taking some thousand filehandles opennms stopped working.
Today a verified the error and built a graph to display which kind of files are ope (lsof -p | group by type). It seems that only the amoung of sockets was increasing. A quick view (lspf -p ... |grep sock) looked like the attached screenshot (lsof.png).
IMHO it looks like the WMI datacollector is not closing its filehandle properly. Maybe someone can acknowledge that
Environment
Operating System: All
Platform: PC
Acceptance / Success Criteria
None
Attachments
2
Lucidchart Diagrams
Activity
Show:
Daniel Hammon October 19, 2010 at 11:38 AM
Damn. I missed to read the release notes Thanks for the fast response. It seems fixed to me, load goes up - open files still down. Cool
Seth Leger (community account) October 18, 2010 at 10:29 AM
An additional fix for this issue went into 1.8.4 after our yearly DevJam developers conference. If you upgrade, this should be fixed. Thanks!
Daniel Hammon October 18, 2010 at 8:49 AM
I imported the group at 12:30
Daniel Hammon October 18, 2010 at 8:48 AM
Created an attachment (id=1109) increasing socket count
I added a provisioning group with 170 nodes and the services icmp and wmi to my setup.
That worked fine over one hour. Data got collected - everything fine.
After one hour opennms needed more and more filehandles and after taking some thousand filehandles opennms stopped working.
Today a verified the error and built a graph to display which kind of files are ope (lsof -p | group by type). It seems that only the amoung of sockets was increasing. A quick view (lspf -p ... |grep sock) looked like the attached screenshot (lsof.png).
IMHO it looks like the WMI datacollector is not closing its filehandle properly.
Maybe someone can acknowledge that