VisionFS daemon using CPU 1

[nashcom]

I'm running VisionFS on OpenServer 5.0.5. It's been working fine since 1999... There are about fifteen client PCs, all Win98. Since yesterday, I've had to reboot the server a few times. It seems that vfsd daemon is taking all the CPU time. It starts off fine, but after a while it increases until the workstations no longer respond, or they can't login.

I've been running Sophos anti-virus software on the workstations, but Sophos have finished support for this version, meaning I have to update to a newer version. I uninstalled the old version from one PC, and installed the new one, and at about the time the server started playing up. It seemed fine after a reboot, and I went ahead and upgraded the other workstations.

How can I work out what's going on with VisionFS?

I know I could try uninstalling Sophos, but even if it is that that's causing the problem, at the minute I can't think how or why. Sophos is set to udate automatically at each PC by downloading straight from the Sophos servers (there are no updates from the SCO/VisionFS server).

David

 

[edube]

Hi, I have a similar problem with SCO 5.0.6.

The vfsd deamon uses all the CPU time but only for 1 user. Every time I kill the vfsd process for that user, the process respawn. The user doesn't know or notice when I kill the vfsd process.

When I run top, I can see many other vfsd process but no one uses all the CPU time like this ?%$ user.

What can I do to diagnose vfsd.

 

[nashcom]

Hi Edube

I 99% certain I've found and fixed my problem. In the VisionFS netlogon directory I had logon scripts for each user - eg david.bat, fred.bat.

Each login script originally had a line in it that checked the version of Sophos on the server, and if the central installation had been updated on the server, it would update the PC. Actually, this was the only line in the login script. When this was no longer necessary dur to the new version of Sophos, I'd commented-out the line using a rem statement.

I think that this rem statement was an invalid syntax - probably on the PC end, and when the login script failed to execute and exit properly, VisionFS got stuck in some sort of loop waiting for that user's login script to finish executing. That's a bit vague, but I think something like that was happening.

I tried deleting the line, such that there was nothing in the login scripts (ie size of 0 bytes) but that also caused the same problem. In the end I just deleted them, and that seemed to fix things.

I used the following commands to check CPU usage

sar -o /tmp/sartmp 30 60

This displays CPU utilisation etc to the screen, taking a snapshot every 30seconds, with a total of 60snapshots.

Also

ps -A -o "pid pcpu ruser args" | grep vfsd

was useful for showing whose process was stealing the show.

Having said that, I think VisionFS had gone belly-up by that time, and I found a reboot of the server was required to sort things.

Let me know if this helps.

David