large NFS copy locks up/hang client with large files (again) (lucid)

[bedge]

This has been a long term problem with Ubuntu, since 10.04. It's an embarrassment and a travesty that they claim enterprise class usability when simple NFS transfers can hang any machine.

[leespa]

Hi All.

For me the main improvement was using 'async' mount options, HOWEVER, I do occasionally experience significant slowdowns when xferring large files...throughput will drop down to nothing and then pick back up to 100+ MB/s again after a few minutes like nothing happened.

I haven't yet determined if this is a RAID-5 or nfs issue...happens sporadically, not always a problem.

[olesk]

I haven't yet determined if this is a RAID-5 or nfs issue...happens sporadically, not always a problem.

I have the same problem using RAID 10 (the real thing, not 1+0) so I don't think it's your RAID5. Upgrading to the latest kernel on my 10.10 solved the crashes however, so this is the only problem that remains for me.

I have seen another oddity though, when copying large movies to my NFS server (a Mac), which is that Ubuntu hangs when trying to generate thumbnails for existing movies (or perhaps of the movie it is transferring) - using 100% CPU and killing pretty much everything. A kill -9 on the process has no effect.

I presume the solution is to disable thumbnails for remote files, but I have already done this to no effect on Nautilus 2.32.0. Nautilus still insists on creating thumbnail on NFS shares - perhaps because it considers them local. If you're having trouble with large files, check if it is movie files that's causing the problem or also ISOs/backups/other large files that freeze, or better yet - try copying from the terminal without having any nautilus windows for the destination open. The problem might actually not be NFS, but Nautilus.

[leespa]

When I see these occasional 'slowdowns' its via rsync in gnome-terminal.

I also noticed copying files to/from same RAID5 volume on the local server is slower than stink...would likely be faster to transfer off and back onto machine.

Thankfully I don't do that operation very often.

[b00nish]

@ olesk

In my case it can't be Nautilus since as KDE user I don't even have it installed. Of course I also tried 'cp' from the terminal and did dozens of other tests.

However since yesterday they finally accepted the bugreport on the Ubuntu bugtracker, so there's hope that the Devs will look at it.

[olesk]

In my case it can't be Nautilus since as KDE user I don't even have it installed. Of course I also tried 'cp' from the terminal and did dozens of other tests.

However since yesterday they finally accepted the bugreport on the Ubuntu bugtracker, so there's hope that the Devs will look at it.

Interestingly, there was an active bug on the system hangs during NFS transfers that now is closed due to a specific fix for this problem in a recent kernel update. If the hangs are still there on the most recent kernel versions, that bug should be reopened (sorry, don't have the link to the relevant bug at my current location).

As for the slowdowns I would like to track that bug as well - do you have the link?

[b00nish]

No, sorry, I don't have any informations related to the slowdown.

(If your system crashes some seconds after the start of the transfer, you'll not notice any suspected slowdown )

All I can say is stated in my Bugreport ( https://bugs.launchpad.net/ubuntu/+bug/661294 ). And this is, that the lock-up occurs with the 10.10 and the 10.04 kernel (2.6.35 resp. 2.6.32) and that they don't occur on my old 8.04 installation. I did not test other kernel versions.

EDIT: Oh, maybe im missunderstood you. If you meant the link to track a bug, it's: https://bugs.launchpad.net/ubuntu/+filebug/?no-redirect
(You'll need a Launchpad-Account, tough)

[nsx241]

This issue has caused me endless grief on 10.04. I was forced to use cifs due to it, along with the painfully slow transfer rates.

However, I recently upgraded to Ubuntu 10.10 (2.6.35.23), and it seems the issue is resolved. I transferred over several 6GB files without any lockup issues, though the transfer rate seems about 10MB/sec slower than before (when it was working correctly on older releases).

[tootlet]

Happening here with Xubuntu 10.10 (and openSUSE 11.3) x86 version. I was running openSUSE 11.2 on this box with 1 320 Gb hard drive with the OS, swap and home partitions, and 3 500 Gb hd's in RAID5 at my business. One of the hard drive broke on it and one failed on my home server at the same time. So I decided to upgrade to 3 1 Tb hard drives and update to openSUSE 11.3. After I got 11.3 set up I tried to rsync some of my data to the new RAID5 via nfs over a gigabyte network. Froze up after about 5 minutes. Hard reboot was only option that worked. On reboot I got an error message stating that no file system was on md0 and to repair it. Couldn't figure that out so after struggling just decided to try Xubuntu 10.10 which was working okay on another server at home. RAID5 built okay and synced. I could make directories. But if I used rsync or cp to copy my Music folder (50 Gb of mp3's) it froze. I would have to resync about 3 hours then try again. Same thing. Requires a hard boot. my /etc/exports file already is async, I didn't try the kernel patch. Rsync has been successful for me many times in the past over this same network. I'll keep googling but suggestions needed. Thanks,
Tom

[nsx241]

Spoke too soon. Problem has returned for me. Until this gets resolved, I'm probably going to switch to Fedora.