CIM Subscription Problems

We have seen some CIM subscription problems when using AlarmVM with ESX hosts running version 4.0. The problem is due to the version of SFCB (Small Footprint CIM Broker 1.3.0 - 1.3.3) that ships with ESX 4.0.

Background

In order for AlarmVM to receive messages about a host's hardware status, it registers itself as a subscriber with the ESX host's SFCB CIMOM broker. During a normal shutdown, AlarmVM unsubscribes from the host and all subscription information is cleaned up. However, if AlarmVM is not gracefully shut down, the subscription information remains in the CIMOM repository. When AlarmVM restarts, it attempts to reuse any existing subscription, but this version of SFCB is often unable to resubscribe. When this happens, AlarmVM will log an error in the vmalarm.log file indicating that we were unable to subscribe to CIM. As a result, AlarmVM will no longer receive any hardware messages from the CIMOM.

Solution

In order to start receiving messages again, we need to clean out the subscription information from the CIMOM repository. 

• Log into the ESX host and change to the SFCB repository directory:
  • /var/lib/sfcb/registration/repository/root/interop

• Edit each of the following files and search for a line with the IP address of the machine that AlarmVM is running on. If you find any lines that contain the IP address, verify that they also contain the string "alarmVM" and then delete them.
  • cim_indicationsubscription.idx
  • cim_listenerdestinationcimxml.idx
  • cim_indicationfilter.idx
• Restart the AlarmVM monitor that is monitoring that host and check the logs for errors. If there aren't any errors in the log, then AlarmVM successfully subscribed to the host's CIMOM.

Note: It may be necessary to restart the SFCB CIMOM service

# /etc/init.d/sfcbd-watchdog restart