intermittent samba hang- needs double restart?

[philjones2000]

Hi

 

My centos file server is the most stable part of my sysem but recently, out of the blue it's started playing up a little.

 

Samba seems to hang. Simply restarting the smb service doesn't do the trick, it needs to be shutdown twice- the first looks as if the smb and nmb services have been shut but the second time another smb service is shut down but not another nmb (I guess it was shut down the first time).

 

Once I've done this there's no probs for a few weeks until the same thing crops up. I'm just worried it'll happen when I'm not around at some point.

 

So what's up? Why the infrequentcy? I haven't changed anything and as there are no Y11 the system is a lot quiter. And why do I have to restart the smb service twice?

 

I don't see anything in the logs- any ideas?

 

Thanks

 

Phil

[Geoff]

It sounds like some of the samba child threads are stuck with zombie status. You can verify this with 'ps'. Now while this might indicate a bug in samba (presumably you've checked for updates) it's more likely this is configuration issue or network related.

 

To troubleshoot this, we need more information. I'd first check how much detail you are logging. Check your 'log level' configuration setting in smb.conf. 1 is almost no logging where as 10 is everything.

 

Also, ensure the samba panic action script is enabled. This will email you whenever one of sambas child processes crashes. You need gdb, sambas debugging symbols and functional email to make this usable though.

[philjones2000]

OK after increasing the logging data and waiting for it to happen I have this to report:

 

Jun 26 14:39:40 byron kernel: KERNEL: assertion (tp->copied_seq == tp->rcv_nxt || (flags & (MSG_PEEK | MSG_TRUNC))) failed at net/ipv4/tcp.c (1348)

Jun 26 14:39:40 byron kernel: KERNEL: assertion (flags & MSG_PEEK) failed at net/ipv4/tcp.c (1284)

Jun 26 14:39:40 byron kernel: recvmsg bug: copied E37FE81B seq E37FE88D

 

Over and over again?

 

Any ideas? Samba is right up to date.

 

Thanks as always.

 

Phil

[Geoff]

Hmm, what network card do you have?

[philjones2000]

It's a intel 82557/8/9 ethernet pro 100...

[Geoff]

Which driver are you using?

[philjones2000]

What's the best way of finding this out? As far as i know i am using the ones that came with fedora... (I got the distro wrong in my first post)

 

Phil

[Geoff]

'lsmod' will tell you what drivers are loaded. It'll be either eepro100 or e100.

[philjones2000]

its the e100

[Geoff]

ok, try the eepro100 driver instead.

[philjones2000]

Thanks for your help Geoff. I'll switch them tonight.

 

Although there is another network card installed on that machine- maybe i should ditch the intel one totally?

[Geoff]

Well I'm suspecting its some sort of TCP offloading problem on the network card. It'd be nice to determine if it is, because you can report it as a bug and have it fixed. Which will benefit everyone.