Files on server suddenly "busy" for no reason?

[Makosuke]

A really strange thing suddenly started happening today at work:

We run about a dozen machines, mostly Macs, that connect to an XServe running Tiger Server for file storage.

Out of nowhere today, several of the Macs (10.4 and 10.3) would, in the middle of working with a file, throw up an error that the user did not have permission to write to the file, or something similar depending on the app.

Closer examination shows that the files are fine, but have somehow been marked as "in use," yet not by the user who was using them. As a result, they cannot be modified (or deleted, except by logging in as an admin user and overriding), although they can be duplicated and the copy worked with just fine.

It's behaving exactly as if the user who was working on the file suddenly changed, leaving the file in "open limbo" and the user unable to do anything to it.

What has me confused is that this problem popped up out of nowhere this morning although NOTHING has changed. In fact, the only unusual thing that's happened in the past month is that all the computers were powered down over the weekend for a scheduled blackout, but everything was fine on Monday so that doesn't seem to explain it.

Has anybody ever seen something like this? Any suggestions as to where to look? I haven't yet found anything even slightly abnormal in the server logs. Gonna be a long night...

 

[SC68Cal]

What protocol are you using to access the file server? AFP?

 

[Makosuke]

AFP. SMB is also running, for the Windows clients, but the Windows machines didn't seem to be having this issue.

I spent the better part of the night reading, double-checking settings, updating anything that theoretically might be related, restarting the server, and so on.

The only thing I changed that could have had an effect was updating the SoftRAID driver from 3.6.2 to 3.6.3, but according to the SoftRAID documentation the only problem with 3.6.2 was with the app, not the driver.

The problem did seem to go away, so my best guess is that another computer came online elsewhere on the network with an IP address that conflicted with the server or the DHCP server screwed up and gave out addresses the same as some of the workstations, and the mess of evening restarts while troubleshooting cleared it up.

We're on a campus network and I have no external control, so that's all I have to point the finger at, but I'd SURE like to know what happened.

Serious problems with mysterious solutions scare me, so any guesses are tremendously appreciated.

 

[SC68Cal]

I know for a fact that AFP can be flaky when the network resource becomes busy, a good example is if you get a disconnect, Finder will come down crashing and burning waiting for that AFP resource, so maybe try switching the mac units over to SMB connections. It'll be easier for you to support anyway.

 

[Makosuke]

Yes, the way Finder (and other apps) reacts to temporarily unavailable network resources is a consistent pain--particularly under 10.3, when computers woke from sleep the delay before the network "woke up" would periodically cause unrecoverable systemwide hangs.

I wasn't aware that this wasn't an issue under SMB, and while I'd rather not switch, since I'm running that anyway I'll give it a shot and see if that helps.

Regardless, though, I'm 99% sure that this isn't the "same" problem, given that I've been running this setup for about three years and this is the first time that's EVER happened.

Perhaps something related, though, so I very much appreciate the suggestion and I'll experiment with it.