SQL Agent ‘Hangs’

[WRShaw]

Hello!

 

We had a delay in our normal backup schedule which resulted in an interesting set of events occur which makes me think there may be a problem with the SQL Agent and PCPDS.EXE.

 

Our nightly backup schedule was running late due to unrelated events that resulted in our SQL backup (on a remote PC) to execute concurrently with the Database Maintenance plans, which in this case were optimizing the database we were backing up.

 

The net result of all of this was a hang of our enterprise backup running on our file server (the SQL server is remote). It appeared that it was waiting for the client to send it more data, but it never timed out.

 

My next thought (since this database cannot be brought down) was to stop the retrospect client service on the SQL Server. That broke the logjam and our backup resumed with the next machine listed in the script (this is not a proactive backup).

 

I restarted the Retrospect client without receiving an error message, but was unable to reconnect using (Configure, Clients). I returned to the SQL server and noticed that Retroclient, RemoteSvc, and PCPDS were all active on the process list. I attempted to stop the Client service but it hung in ‘stopping’. It wasn’t until many hours later when I could cycle the SQLServer service and idle some of our workforce did PCPDS stop with an error. I started the client, received the message the Retrospect Client is already running. I checked the process list and found two instances were running, so I stopped the service again and restarted and all was well in the world.

 

I think we have a couple of opportunities here:

 

1). The PCPDS task should have detected there was an issue with the database condition and returned an error to Retrospect.

2). Failing that, one client computer shouldn’t be able to hold up a backup script indefinitely.

3). Failing both of those and in case of manual intervention I think we should have the ability to stop (kill if necessary) the retrospect agents to prevent cycling the server when stopping a task is all that is really needed.

 

Any thoughts/comments???

 

Thanks,

Bill