Unsolved

This post is more than 5 years old

1 Message

1049

December 18th, 2006 02:00

Backbone and Agent services are terminated unexpectedly

I've several installations of Autostart 5.2.1 on Windows server 2003 SP1.
On one server the Backbone service terminate unexpectedly very often (1 to 2 times per hour) but only when a resource group is running on it. If the same resource group is running on the other node everything goes well.
Each time the problem occurs I can see an error (Event ID 7034) in the event log of this server. The Agent service then fails because of the service dependency. That generated no failover. The two services are automatically restarted 20 seconds after the failures occurred but I don't know why as there is no recovery action specified for these services in Windows.
I've already try to:
1) reinstall Autostart on this node
2) completely uninstall and reinstall the cluster
3) checked the antivirus software. EMC Autostart folder is excluded from scan.
The problem is the same. It is very strange because I use the same hardware, same revision of OS and drivers and all servers and I have other Autostart installations working fine.
Could someone have an idea?
Thank you
Emmanuel

198 Posts

December 18th, 2006 03:00

Have you tried enabling full AutoStart logging of all the components on this node?

1 Message

January 29th, 2007 08:00

How can you resolve this problem?
I have the same problem.

45 Posts

July 30th, 2007 13:00

This is almost always due to a broadcast issue on your domain lines. On EVERY install you should ALWAYS configure point-to-point for all of your domain lines. The hardware today "should" handle the broadcast but it doesn't.

I know it is a pain. I had an install where I did 16 HPUX servers with 4 NICs each!!! One path from each NIC to all other NIC's on all servers. You do the math. However, I tried going without point to point because I really didn't want to create all those domain lines but I had these types of problems where the Agents would go down all the time for no reason. After setting everything to point to point it worked without issues.

Just depends on the customer's hardware. The past two years, I have ALWAYS defined the domain lines as point to point and haven't seen this problem.

0 events found

No Events found!

Top