Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
[Holger Levsen] Hi Wolfgang, thanks for confirming! Yes. Great to have it confirmed that there is a race condition involving nslcd, autofs and network-manager. The fact that it is hidden when network-manager is quicker is not really comforting, as it can trigger at any time if there is a slow dhcp server or some other problem related to the network. Did you manage to figure out if nslcd could be told to wait until the network is up? Did you try to restart it in the dhclient hook when the network is up? I suspect the latter is required to get a robust setup when machines change IP address, and the former is required to ensure nslcd always work after boot. -- Happy hacking Petter Reinholdtsen -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Tue, Dec 16, 2014 at 12:21:06PM +0100, Petter Reinholdtsen wrote: Did you manage to figure out if nslcd could be told to wait until the network is up? No. But there's already /etc/network/if-up.d/nslcd. This script does nothing, if nslcd died before. But may be it could check if the service should have been started and start it in this case. As far as I found out, systemd has a target called 'network-online'. Maybe this could be used within a nslcd systemd service file. But I don't know if this would really solve the problem and I don't know enough about systemd. Did you try to restart it in the dhclient hook when the network is up? Yes: nslcd-restart, more or less copy + sed from autofs-reload. This works like expected _now_. But it caused the system to hang at the stage 'raising network interface' in the case of failing NFS mount; seemed to be a dead end street for me. I had to use 'init=/bin/sh' on the kernel command line to be able to remove the hook script. Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Tue, Dec 16, 2014 at 04:08:12PM +0100, Wolfgang Schweer wrote: As far as I found out, systemd has a target called 'network-online'. Maybe this could be used within a nslcd systemd service file. But I don't know if this would really solve the problem and I don't know enough about systemd. Seems to be that ordering is already like that, just look at the output of these commands: systemctl show network-online.target and systemctl show nslcd.service Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
Hi Wolfgang, (I guess my question is pretty obvious, but I want to be sure...) On Sonntag, 14. Dezember 2014, Wolfgang Schweer wrote: Added a default (desktop=kde) workstation: NFS mount fails all the time. this is due to the old (=not from sid) network-manager being installed and used? But: NFS mount seems to work reliably with network-manager v. 0.9.10.0-4 (unstable) installed. The network interface is now brought up pretty fast (took up to about 30 seconds before). Maybe now the right sequence of services is given. nice! Thanks for your continous work on this! Totally awesome! cheers, Holger signature.asc Description: This is a digitally signed message part.
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Mon, Dec 15, 2014 at 01:03:50PM +0100, Holger Levsen wrote: this is due to the old (=not from sid) network-manager being installed and used? Yes, atm version 0.9.10.0-3; network-manager v. 0.9.10.0-4 was the only package installed from unstable to make NFS mount work. Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
Hi Wolfgang, thanks for confirming! On Montag, 15. Dezember 2014, Wolfgang Schweer wrote: Yes, atm version 0.9.10.0-3; network-manager v. 0.9.10.0-4 was the only package installed from unstable to make NFS mount work. ok, cool, 0.9.10.0-4 should migrate to jessie in four days: https://packages.qa.debian.org/n/network-manager.html says Too young, only 1 of 5 days old. Ignoring block request by freeze, due to unblock request by nthykier. cheers, Holger signature.asc Description: This is a digitally signed message part.
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Sun, Dec 14, 2014 at 01:38:22AM +0100, Wolfgang Schweer wrote: Strange enough, with a totally stripped down VM test environment NFS mount succeeds all the time without any changes to rc.local and without adding a dhcp hook file. This is the setup: main server (with 'desktop=kde' removed from the kernel command line) and only the main-server profile. 'apt-get install xterm iceweasel icewm lightdm' to get a minimal gui to make GOsa² available. client installation with profile 'minimal', system added using GOsa². To be continued... Added a default (desktop=kde) workstation: NFS mount fails all the time. But: NFS mount seems to work reliably with network-manager v. 0.9.10.0-4 (unstable) installed. The network interface is now brought up pretty fast (took up to about 30 seconds before). Maybe now the right sequence of services is given. Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Thu, Dec 11, 2014 at 08:40:21PM +0100, Wolfgang Schweer wrote: Then I changed the commands in rc.local like this: # By default this script does nothing. systemctl stop nslcd systemctl start nslcd systemctl restart autofs exit 0 With this workaround NFS mount succeeded 10 out of 10 times. I've tested these commands in a dhcp client exit hook file two days ago and that worked as well. Sad to report that after upgrading server and client today NFS mount fails all the time. Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Sun, Dec 14, 2014 at 12:07:17AM +0100, Wolfgang Schweer wrote: Sad to report that after upgrading server and client today NFS mount fails all the time. Strange enough, with a totally stripped down VM test environment NFS mount succeeds all the time without any changes to rc.local and without adding a dhcp hook file. This is the setup: main server (with 'desktop=kde' removed from the kernel command line) and only the main-server profile. 'apt-get install xterm iceweasel icewm lightdm' to get a minimal gui to make GOsa² available. client installation with profile 'minimal', system added using GOsa². To be continued... Wolfgang signature.asc Description: Digital signature
Bug#759544: Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
tags 759544 + help thanks On Wed, 2014-12-10 at 22:32 +0100, Holger Levsen wrote: Raising the severity of 759544 as 20% failure rate surely has a major impact on the usuability of the package... Please see 772342 to learn how 759544 affects debian-edu-config. Any help in tracking this down is highly appreciated. Since I've switched to putting network configuration into /etc/network/interfaces and ignoring network-manager my setup with NFS and nslcd is not failing during boot. I've seen two issues with nslcd start-up: - one that logs unable to daemonize in some cases (the patches mentioned earlier in this bug report may fix this) - one when using TLS (also see https://bugs.debian.org/643948) Also nslcd is supposed to be started after networking has been brought up (LSB headers) but this doesn't seem to work in all cases. Thanks, -- -- arthur - adej...@debian.org - http://people.debian.org/~adejong -- signature.asc Description: This is a digitally signed message part
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Thu, Dec 11, 2014 at 12:36:30AM +0100, Wolfgang Schweer wrote: [ /etc/rc.local ] # By default this script does nothing. systemctl restart nfs-common systemctl restart autofs systemctl stop nslcd systemctl start nslcd exit 0 Investigating further (virtual-box VM setup) I dropped the connection to the internet gateway, then restarted main server and workstation. With the above stated commands in rc.local the workstation failed 6 out of 6 times to get NFS mount right. Then I changed the commands in rc.local like this: # By default this script does nothing. systemctl stop nslcd systemctl start nslcd systemctl restart autofs exit 0 With this workaround NFS mount succeeded 10 out of 10 times. So I guess the NFS issue boils down to the fact that order of services' start matters. Maybe a proper sequence can be achieved using nslcd and autofs systemd services files containing the right targets. Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Sat, Dec 06, 2014 at 02:58:29PM +0100, Wolfgang Schweer wrote: Package: debian-edu-config Version: 1.814 Severity: important NFS mount from tjener seems to work only ramdomly. This happens even after 'autofs reload' was added to dhclient exit hooks. Once I got it working with autofs stop/start instead of reload. Seems to be that in addition to autofs nslcd is playing only randomly right, too. After a new installation of a main server and a workstation (installed via PXE using d-i-8.0 daily netboot.tar.gz) NFS mount worked out of the box yesterday -- but failed today. 'systemctl status nslcd.service -l' revealed that nslcd failed to daemonize. After stopping/starting services NFS mount was working again: systemctl stop nfs-common; systemctl start nfs-common; systemctl stop nslcd; systemctl start nslcd; systemctl stop autofs; systemctl start autofs Seems to be that #759544 is tracking this issue. Wolfgang signature.asc Description: Digital signature
Bug#759544: Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
severity 759544 important affects 759544 debian-edu-config thanks Raising the severity of 759544 as 20% failure rate surely has a major impact on the usuability of the package... Please see 772342 to learn how 759544 affects debian-edu-config. Thanks. signature.asc Description: This is a digitally signed message part.
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
On Wed, Dec 10, 2014 at 10:23:35PM +0100, Wolfgang Schweer wrote: On Sat, Dec 06, 2014 at 02:58:29PM +0100, Wolfgang Schweer wrote: NFS mount from tjener seems to work only ramdomly. This happens even after 'autofs reload' was added to dhclient exit hooks. Seems to be that in addition to autofs nslcd is playing only randomly right, too. After a new installation of a main server and a workstation (installed via PXE using d-i-8.0 daily netboot.tar.gz) NFS mount worked out of the box yesterday -- but failed today. Fed up with the randomness, I added some commands to the client's /etc/rc.local file as a workaround to be able to test the system. #!/bin/sh -e # # rc.local # # This script is executed at the end of each multiuser runlevel. # Make sure that the script will exit 0 on success or any other # value on error. # # In order to enable or disable this script just change the execution # bits. # # By default this script does nothing. systemctl restart nfs-common systemctl restart autofs systemctl stop nslcd systemctl start nslcd exit 0 I rebooted the client two times and controlled /var/log/syslog each time. The first time nslcd failed to daemonize but NFS mounting home worked due to nslcd being started from rc.local; second time the workaround wouldn't have been needed cause nslcd behaved nicely from the beginning. Wolfgang signature.asc Description: Digital signature
Bug#772342: debian-edu-config: mounting homedirs via NFS doesn't work reliably
Package: debian-edu-config Version: 1.814 Severity: important NFS mount from tjener seems to work only ramdomly. This happens even after 'autofs reload' was added to dhclient exit hooks. Once I got it working with autofs stop/start instead of reload. Maybe there's another race condition causing this issue? Wolfgang signature.asc Description: Digital signature