new vzdump behavior

[proxmox@secu.be]

Hello
I'm just finished to migrate a cluster from proxmox 1.9 to 2.0 with 99.9 % success

1) on some server, apache did'nt restart after joining a cluster because of a missing cert
I need to do "pvecm updatecerts" manually to solve the problem

2) vzdump backup create new problems not existing in the previous version
I can see these errors in the log
rsync: readdir("/mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp/usr/share/man/man8"): Too many levels of symbolic links (40)
INFO: IO error encountered -- skipping file deletion
INFO: Number of files: 355717
......
INFO: rsync error: some files/attrs were not transferred (see previous errors) (code 23) at main.c(1060) [sender=3.0.7]
INFO: resume vm
INFO: Resuming...
INFO: vm is online again after 103 seconds
cannot remove directory for /mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp/usr/share/man/man8: Directory not empty at /usr/share/perl5/PVE/VZDump.pm line 962
cannot remove directory for /mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp/usr/share/man: Directory not empty at /usr/share/perl5/PVE/VZDump.pm line 962
cannot remove directory for /mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp/usr/share: Directory not empty at /usr/share/perl5/PVE/VZDump.pm line 962
cannot remove directory for /mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp/usr: Directory not empty at /usr/share/perl5/PVE/VZDump.pm line 962
cannot remove directory for /mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp: Directory not empty at /usr/share/perl5/PVE/VZDump.pm line 962
ERROR: Backup of VM 1101 failed - command 'rsync --stats -x --numeric-ids -aH --delete --no-whole-file --inplace '/var/lib/vz/private/1101/' '/mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp'' failed: exit code 23
INFO: Starting Backup of VM 111 (openvz)
INFO: CTID 111 exist mounted running
INFO: status = running
INFO: backup mode: suspend
INFO: ionice priority: 7
INFO: starting first sync /var/lib/vz/private/111/ to /mnt/pve/nfs1/dump/vzdump-openvz-111-2012_04_11-04_07_13.tmp

any ideas how to solve that ?

3) in the gui, one server is displayed in red instead of green since yesterday
I can see that cman is not started
If I try to start it to start it again via the gui ; I got the following error message
Waiting for quorum...
[FAILED]
TASK ERROR: command '/etc/init.d/cman start' failed: exit code 1
but I'm 100% sure that the config is exactly the same than the other servers


Thanks anyway because it was incredible to migrate everything in just an eastern holiday week end ! Keep on working

4) anyone working with the json api to provide a gateway to whmcs ? if yes, do you need developer or beta tester ?

 

[dietmar]

2) vzdump backup create new problems not existing in the previous version
I can see these errors in the log
rsync: readdir("/mnt/pve/nfs1/dump/vzdump-openvz-1101-2012_04_11-03_24_03.tmp/usr/share/man/man8"): Too many levels of symbolic links (40)

Never saw that. Can you manually rsync that container?

 

[proxmox@secu.be]

Hello
My openvz VM is the result of a rsync since I used the rsync method to convert a physical server to a virtual one
And rsync is working well in the proxmox VE 1.9 suspend and vzdump process

I remember of a problem I've got once with "rsync --rLptgoD --delete source dest" (big L means copy symlinks as files ) and "rsync --rlptgoD --delete source dest" (small l means copy the symlinks)

 

[proxmox@secu.be]

[h=5]Re: vzdump issue[/h] « Reply #3 on: January 22, 2012, 11:44:55 AM »
Ok, it seems to be caused by a buggy tar version 1.23 (it fails to remove symlinks). Upgrading should help the issue. I've replicated the condition and testing it out currently.

 

[marco44]

I have exactly the same error (with a NFS mount).

In the dmesg, I get NFS: directory man/man7 contains a readdir loop. Please contact your server vendor. The file: pam.7.gz has duplicate cookie 599672783. Feels like a NFS bug to me. There as been a lot of them corrected lately…