We recently started having a lot of problems with our CPS-protected File server. Windows was telling users that the Fileserver disk was full whenever they tried to save anything to it despite the server actually having Gigabytes of free space. The server itself was reporting the following 2021 Errors in the system log.

Event Type:         Warning
Event Source:     Srv
Event Category:                None
Event ID:              2021
Date:                     6/02/2008
Time:                     4:51:40 PM
User:                     N/A
Computer:          FS
Description:
The server was unable to allocate a work item 9894 times in the last 60 seconds.

The only way to fix the issue was to reboot the server. Our interim solution was to schedule a server reboot early every morning in order to minimise this problem. Obviously this was not an acceptable long term solution. 

We’ve spent a lot of time with this KB article:
http://support.microsoft.com/kb/317249

We’ve upped our MaxWorkItems and tweaked SizReqBuf.
We’ve adjusted the settings in HKLM\SYSTEM\CurrentControlSet\Services\lanmanserver\parameters, just incase.
We’ve uninstalled RAWS and disabled the SAV 10.2 realtime AV on the server.
Most of the information I can find on srv event 2021 points to one of the things above, or to a kernel filter driver causing the problem.

The problem first started not long after we installed CPS, so we’ve disabled CPS’s filter driver.
http://support.microsoft.com/kb/816071
The specific service we disabled was CpsFsJnl
HKLM\System\CurrentControlSet\Services\CpsFsJnl
We changed the value of the “Start” key from 2 to 4.

With the CPS filter driver disabled, we also switched off our daily reboot job and got 8 straight days without a crash or reboot. With the filter driver enabled again we started getting 2021 errors in around 12 hours.

We’re using BackupExec 11D build 7170.
RAWS is not installed on the machine that’s crashing (was originally, was removed as part of the troubleshooting).
The version of CPS is 11.1.407.4003 and the CpsFsJnl.sys file is version 11.0.400.701, build number 4.00.07a. Modified date is 6 Dec 2006.
The server in question is running Windows Server 2003 R2 Enterprise i386 with SP2.
It’s a file server and AD DC, and is running DFS Replication to a few other sites.
The majority of the files on this server live on a LUN on our EMC CX3-10 SAN, connected by FC. The disks are 15KRPM and the LUN is a RAID5 set.
We have about 70 workstations using this server as their main FS.

Has anyone else out there had problems with CPS knocking over servers under load, or causing 2021 events?
Is there a more recent version of the CpsFsJnl.sys driver and if so, where can I get it?
Has anyone here migrated from BE CPS to Microsoft DPS and if so, how would you compare the two?

Ryan.