NFS file transfer freezes the NFS client machine.

[Gulam Mohiuddin]

NFS file transfer process freezes up the NFS client machine. Although about 30GB data get copied eventually. But during this transfer period we can not access the target (NFS client) machine. The source (NFS Server) is fine during copying process.

What could be the reason causing NFS client machine to freeze up?

Note: Bothe NFS Client & NFS Server are HP-UX 11i boxes (LAN).

Following is my NFS export command from the NFS Server:
exportfs -i -o ro /u27/oracle /u28/oracle /u29/oracle
exportfs –v


Also anything I can monitor during this NFS copy process?



Thanks,

Gulam.

[Steven E. Protter]

Shalom Gulam,

Check /var/adm/syslog/syslog.log on the client for problems at the time this is happening. The problem could be networking, full fileystems, nearly anything. Lets try and narrow it down.

SEP

[Gulam Mohiuddin]

I tried again, but there is no error or messages reported in syslog.log.

Is there any NFS log?

Anything else could give us some idea?


Thanks,

Gulam.

[MarkSyder]

Are the network cards running on different speeds? If the card on the receiving system is slower that could cause a problem like this.

Mark Syder (like the drink but spelt different)

[Gulam Mohiuddin]

NIC are same on the both the servers.

This NFS copy process was working fine for more than one year, but suddenly from last month it started freezing up the NFS client machine completely.


Thanks,

Gulam.

[MarkSyder]

Disc space on the receiving machine?If it was working for over a year then suddenly wasn't, perhaps the filesystem is close to being full.

Mark

[Gulam Mohiuddin]

All filesystems are less than 90% used after completion of NFS transfer, will it be an issue even with 10% free space?


Thanks,

gulam.

[MarkSyder]

What I would suggest is to log in to the receiving server and run the command top. Then start a transfer and see what the CPU use is; especially that of the top process. If it will let you run commands, you might also try swapinfo -tam in case it's a memory issue.

Mark

[Gulam Mohiuddin]

The receiving server freez up after I start the NFS process, so I can not open any new session or old sessions.

Thanks,

Gulam.

[MarkSyder]

How about if you start top running before you start the transfer?

I use Exceed from a PC, so swapping from one server to another is simply a matter of clicking on a tab. Do you not have a similar setup?

Mark

[Gulam Mohiuddin]

I have the similiar setup.

But after starting NFS process I can not open even previously started Exceed Sessions.

Everything just freezes up, no matter even if you start Top command in another Window before starting NFS process.

Thanks,

Gulam.

[Dave Olker]

Hi Gulam,

How about starting up a very high priority shell and running top/glance from that shell? Maybe then it could get CPU time enough to show you where the problem lies?

Dave

[Gulam Mohiuddin]

Good suggestion, but please let me know how to do that?

Thanks,

Gulam.

[Dave Olker]

Here's how I usually do it:

# rtprio 127 ksh

Just be very careful with this shell since you're going to be running at a very high priority and not getting preempted by much else so if you're not careful you can wreak a lot of havoc on the system.

Regards,

Dave

[Gulam Mohiuddin]

OK, just to make sure how would I stop it. I guess just use the "exit" command.

Gulam.

[Dave Olker]

Yes, just exit the shell just like any other shell.

[Gulam Mohiuddin]

I tried using rtprio 127 ksh prior to starting my NFS process.

But unfortunately still everything is freezed up, I can not see any session even which I started up with "rtprio".

Please help.

Gulam.

[Gulam Mohiuddin]

Now I have lost all three Exceed connection which I have started prior to running my NFS Job.

Thanks,

Gulam

[Dave Olker]

Are you able to log back into the client or is the system completely hung? If it's completely hung and non-responsive then force a TOC dump and have someone in HP support look at the dump. If you can get back on the client, I would be curious if things change if you unmount and re-mount the filesystems before firing off another test.

Regarding an earlier post - yes, things can get bad even if there is 10% free disk space. A lot depends on how fragmented the filesystem is, because if you're doing a large data transfer the chances are you're going to need a lot of *continguous* chunks to store the file. If your filesystem is already near 90% full and the remaining 10% is heavily fragmented then the system will have to work extremely hard to build the extents needed to hold the incoming data.

When is the last time you defragmented your filesystem? Can you either increase the size of the filesystem or remove some files from it and then defrag it to see if having more contiguous space in the filesystem helps reduce the CPU overhead?

Regards,

Dave

[Gulam Mohiuddin]

No, I can not log back to the client or can not open new Exceed sessions.

Everything is just frozen up.

Gulam.

[Dave Olker]

Ok, well then you can either wait for things to clear up (assuming they will after the transfer completes) or you can get onto the console of the system and force a TOC dump.

Then I'd be curious about the rest of the questions I posed in my previous post.

Dave

[Gulam Mohiuddin]

My NFS job completed successfully.

And everything else is back to NORMAL.

Althoug, no clue yet why the server getting freez up during NFS transfer.

Thanks,

gulam.

[Dennis Handly]

I've had "hangs" while a core file gets dumped.
All accesses in that directory are hung. It seems the kernel doesn't realize that the priority should be very low and let other traffic go through.

I'm not sure if your NFS case is similar?