[Bug 420575] New: System Freeze and NFS hang when switching network connections
https://bugzilla.novell.com/show_bug.cgi?id=420575 Summary: System Freeze and NFS hang when switching network connections Product: openSUSE 11.1 Version: Alpha 2 Platform: Other OS/Version: openSUSE 11.0 Status: NEW Severity: Normal Priority: P5 - None Component: KDE3 AssignedTo: kde-maintainers@suse.de ReportedBy: quentin.jackson@exclamation.co.nz QAContact: qa@suse.de Found By: Customer This applies to opensuse 11.0 and 11.1 alpha 1/2. When undocking or redocking from a docking station or doing the same via cabled network into laptop, the system will freeze, slow down and NFS mounts are no longer accessible. Doing an /etc/init.d/nfs restart does not fix the problem, restarting the computer does. I've also noticed unplugging a network cable on a desktop can cause the machine to freeze. The knetwork manager does connect to the alternate connection (wireless or desktop), firefox does work, but the kicker button will not work and NFS mounts will not work. I'm surprised no one else has found this, as it's very annoying. :) I'm running KDE3 series, latest alpha, but as above is also present on 11.0 GM. If there's any info anyone needs it's easy to reproduce 100% of the time :) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User jpr@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c1 JP Rosevear <jpr@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|kde-maintainers@suse.de |nfbrown@novell.com Component|KDE3 |Basesystem --- Comment #1 from JP Rosevear <jpr@novell.com> 2008-09-18 11:33:22 MDT --- Well, I think this is a general nfs issue with networking going down, the nfs maintainer can probably give more background, but I'm not sure its easy to fix. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User quentin.jackson@exclamation.co.nz added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c2 Quentin Jackson <quentin.jackson@exclamation.co.nz> changed: What |Removed |Added ---------------------------------------------------------------------------- Severity|Normal |Major --- Comment #2 from Quentin Jackson <quentin.jackson@exclamation.co.nz> 2008-09-18 13:54:03 MDT --- I'm sure that NFS does have a problem, I'm just not sure that it's only NFS, I've found the new network manager to be quite different than in the 10 series in the way it behaves, in fact I wouldn't be surprised if the way it now handles dual NIC's has caused a problem in NFS. In any case if there are any logs I can post to show what the problem actually is, please do as this is a major problem for anyone using NFS. I'm raising Priority on this as normal priority things don't seem to get attended to quite the same and I believe this is Major :) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c3 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|nfbrown@novell.com |bnc-team-screening@forge.provo.novell.com --- Comment #3 from Neil Brown <nfbrown@novell.com> 2008-09-21 18:32:48 MDT --- NFS doesn't cope with the client's IP address changing. One option is to use UDP for mounting rather than TCP (-o udp). In this configuration, changing the client address causes fewer problems. However there are other problems with using UDP, such as a small possibility of data corruption, particularly at high data rates. In either case, if you hold any locks on NFS files before changing IP address, you probably will having significant problems accessing those files after the change. Probably the best approach would be to have all NFS filesystems mounted by an auto-mounted, and have Network Manager unmount any NFS filesystems before changing the IP address. This would need to be a --lazy unmount in case some process were using the filesystem. That process would probably hang, but any other process that tried to access the filesystem would we a new mount. I think that it would always be a good idea for NetworkManager to unmount --lazy any NFS filesystem when changing IP address, because they will be completely useless and can be harmful. This should not include NFS filesystems that are mounted via the loopback address (127.0.0.1) or top-level mounts of the auto-mounter. As a longer term response I will look into the possibility of enabling an NFS mount to move smoothly to a new client IP address, but at this stage I think that situation is best handed by NetworkManager. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User quentin.jackson@exclamation.co.nz added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c4 --- Comment #4 from Quentin Jackson <quentin.jackson@exclamation.co.nz> 2008-09-21 22:38:22 MDT --- That's good, but I think something else has changed though. I never used to have this problem on 10.3. Also it stops the computer booting for quite a while 1-2mins if the NFS server isn't there or the network is down which it never used to do on 10.3. Quite honestly when I first started using NFS file shares I was expecting this behaviour on 10.3 and was pleasantly surprised when it wasn't a problem. When we start getting symptoms like the KDE Kicker button freezing as a result it becomes quite apparent that something bigger is going on, (it would seem to me at least). * So in short, 10.3 doesn't hang when NFS server cannot be found 11.0 does. * 10.3 didn't have problems switching between the wireless network and the wired network using network manager, 11.0 does. * I've also noticed unplugging the NIC cable on 11 can cause the machine to hang. So are we sure this is not a regression? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c5 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |nfbrown@novell.com --- Comment #5 from Neil Brown <nfbrown@novell.com> 2008-09-24 22:29:58 MDT --- What filesystem(s) do you have mounted over NFS? Are they mounted via /etc/fstab or by and automounter. What mount options are used? If an NFS filesystem is mounted in /etc/fstab without the 'bg' option, I would expect a delay at boot-time. There shouldn't be a delay if you have 'bg'. Switching from wired to wireless would work if you used UDP, but I don't think it would ever work with TCP. I don't think 10.3 defaulted to UDP. I don't suppose you still have your 10.3 installation that you can check? Unplugging the NIC cable would only cause a hang when something accessed the filesystem. Maybe something is accessing filesystems in 11.0 in a way that wasn't happening in 10.3. Maybe the answers to the above three questions will help make the situation clearer. Looking forward: if you want NFS to work at all reliably while your laptop changes IP address, you would really need to set up something like OpenVPN so your laptop can have a stable address (on the VPN) while it changes network. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 Christoph Thiel <cthiel@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|bnc-team-screening@forge.provo.novell.com |nfbrown@novell.com -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User quentin.jackson@exclamation.co.nz added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c6 --- Comment #6 from Quentin Jackson <quentin.jackson@exclamation.co.nz> 2008-09-26 16:15:43 MDT --- I have xfs filesystems on a software raid 5 mounted over NFS, they are set up by the NFS client tool in YaST and effectively end up in /etc/fstab therefore. No idea how else to do it :) The entries copied from fstab are: 192.168.1.10:/home/username /home/quentinj/NFS/username nfs defaults 0 0 192.168.1.10:/mnt/data1/Video /home/quentinj/NFS/Video nfs defaults 0 0 192.168.1.10:/mnt/data1/Audio /home/quentinj/NFS/Audio nfs defaults 0 0 192.168.1.10:/mnt/data1/MyPics /home/quentinj/NFS/MyPics nfs defaults 0 0 Hadn't heard of the bg option before, should that be added so that YaST adds it automatically here or does it cause other side effects? I have never knowingly used UDP or TCP specifically for NFS and I don't have the 10.3 any more sorry, though I would install it if I wasn't so scared I'd get burnt by the beta 1 grub problem that I was lucky to get around! :) What I do know is that there were quite a few new problems with NFS in OpenSuSE 11.0 and to date I still turn off NFS 4 options as I cannot get a connection with them, so I still find the NFS stuff suspicious. I also notice interesting behaviour with the network manager in that nowadays it maintains 2 simultaneous connections rather than switching between the two and I don't think it updates the correct route to send traffic through. Wireless doesn't even work for me since beta 1 so I can't test it any longer anyway. I just tried replicating the unplugging of the NIC card and the PC didn't crash, it used to on 11.0 (a seperate machine I've got) so I guess that's been fixed via network update. Looking forward no one is going to set up a vpn to get NFS working sorry. The suggestion higher up to have network manager unmount and remount sounds more doable though. Hey, windows can do it, Linux should be able to do it much easier! ;) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c7 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |quentin.jackson@exclamation.co.nz --- Comment #7 from Neil Brown <nfbrown@novell.com> 2008-12-04 14:47:25 MST --- I think the network manager has now be modified to unmount NFS filesystems at disconnect and remount them at reconnect. If you could test -RC1 or the latest Factory, that would be great. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User quentin.jackson@exclamation.co.nz added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c8 Quentin Jackson <quentin.jackson@exclamation.co.nz> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|quentin.jackson@exclamation.co.nz | --- Comment #8 from Quentin Jackson <quentin.jackson@exclamation.co.nz> 2008-12-04 16:36:18 MST --- Thanks, it kind of works. I don't get all the horrible system freezes any more, but when switching from wired to wireless it doesn't remount, I have to do that manually. So it appears we're half way there, but certainly a big improvement, thanks! Q -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c9 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |tambet@novell.com --- Comment #9 from Neil Brown <nfbrown@novell.com> 2008-12-04 18:30:03 MST --- Hi Tambet, Could you please comment on whether you would expect network manager to remount NFS filesystems when the network comes back up. Probably running rcnfs restart would be a sensible thing to do. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User tambet@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c10 Tambet Ingo <tambet@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |tambet@novell.com Status|NEEDINFO |NEW Info Provider|tambet@novell.com | --- Comment #10 from Tambet Ingo <tambet@novell.com> 2008-12-05 01:17:38 MST --- That's what it already does: /etc/init.d/nfs restart -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c11 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |quentin.jackson@exclamation.co.nz --- Comment #11 from Neil Brown <nfbrown@novell.com> 2008-12-07 15:00:21 MST --- Thanks. So it looks like things are configured the right way so the mounts should be restored when you switch networks, but it seems they aren't. Quentin: Could you please try again to switch from one network to the other and make sure there is at least a 1 minute gape between disabling the first and enabling the second network. See if that makes a difference. If it doesn't - i.e. after switching networks the nfs mounts aren't mounted, try running /etc/init.d/nfs restart and see if that makes the mounts work. Also see if any errors are reported either on the screen or in a log file (/var/log/messages maybe). Let me know the results. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User quentin.jackson@exclamation.co.nz added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c12 Quentin Jackson <quentin.jackson@exclamation.co.nz> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|quentin.jackson@exclamation.co.nz | --- Comment #12 from Quentin Jackson <quentin.jackson@exclamation.co.nz> 2008-12-08 15:06:55 MST --- Great timing :) I actually did this just this morning before getting on the plane. The thing is that it's between switching wireless and wired which makes it impossible to put a delay in as it is done automatically. Restarting the NFS service definately fixes it. I think we're on the same wavelength cause' this morning I was thinking it needed a slight delay before restarting the services. It seems to take about 5-10 seconds for either wireless or wired network to come up, so maybe we can delay the NFS restart a little? Or maybe it's not running after connection confimation is obtained from network manager? Q -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User tambet@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c13 --- Comment #13 from Tambet Ingo <tambet@novell.com> 2008-12-08 23:43:44 MST --- The NM dispatcher scripts are run after the devices are up and fully configured. It is very easy to confirm the dispatcher scripts run and that the network connection is really working - Add something like: date >> /tmp/test ping -c 1 $your_gateway_address >> /tmp/test to /etc/NetworkManager/dispatcher.d/nfs. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c14 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |quentin.jackson@exclamation.co.nz --- Comment #14 from Neil Brown <nfbrown@novell.com> 2008-12-10 16:14:06 MST --- Setting to NEEDINFO - see previous comment. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
https://bugzilla.novell.com/show_bug.cgi?id=420575 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=420575#c15 Neil Brown <nfbrown@novell.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |CLOSED Info Provider|quentin.jackson@exclamation | |.co.nz | Resolution| |WORKSFORME --- Comment #15 from Neil Brown <nfbrown@novell.com> 2009-03-04 21:52:36 MST --- Closing due to inactivity. Please reopen if new information becomes available. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
participants (1)
-
bugzilla_noreply＠novell.com