In the past few weeks I have run into (3) issues with SRM polling devices. The first is the SNMP daemon crashing and restarting on MDS and Nexus switches, the next was all of the SNMP daemons on the Nexus switches stopped requiring Cisco to install a temp patch on the system that gave them access to the Linux kernel in order to reset the service (we ended up having to reboot 1 switch which was a disruption to our fabric). The last happened today with a VPlex management server where if you typed 1 or 2 commands the java process ran out f memory. I did not see any advisory on this regarding changing any of the polling information. I think the problem is how often or how much information SRM tries to gather at one time. On the MDS and Nexus switches I am narrowing it down to the SNMPGetBulk request. I have turned this off and have not seen any SNMPd crashes. On the VPlexes it was a max sessions command changed from the default of 10 to 20. Can anyone telll me if there is a document addressing this? We are in the beginning stages of upgrading to ViPR SRM 3.5. Thanks.
Paula, below are some of the
knowledgebase articles on ViPR SRM "Support by Product" site.
SRM Suite 3.0: Cannot connect to VPLEX Management
Station because SRM is polling VPLEX too frequently; How can we change the SRM
VPLEX SolutionPack collection polling interval?
ViPR SRM/Watch4Net: Performance of Cisco MDS switches
is impacted by ViPR SRM 3.0
Both articles 000186813 (VNX) and 000184791 (VPLEX) seem to be addressed as of ViPR SRM 3.0.2. In the case of VPLEX, changes were made to limit simultaneous requests from ViPR SRM to the VPLEX management station. In addition, there is a patch available for VPLEX 5.2 and 5.3 to fix resource leaks that should also be applied.