Snippet pre-start runs during backup

[Elliott Partridge]

I have a VM that has a snippet configured with pre-start and post-stop actions. The VM is currently stopped. When my backup job runs (which includes all VMs & containers), the pre-start action in the snippet is executed. Is this intentional? How do I distinguish between pre-start due to backup vs actually starting the VM/container?

INFO: Starting Backup of VM 104 (qemu)
INFO: Backup started at 2021-04-06 22:13:45
INFO: status = stopped
INFO: backup mode: stop
INFO: ionice priority: 7
INFO: VM Name: pbs1-vm
INFO: include disk 'scsi0' 'vm-prod:vm-104-disk-0' 8G
INFO: creating vzdump archive '/var/lib/vz/dump/vzdump-qemu-104-2021_04_06-22_13_45.vma.zst'
INFO: starting kvm to execute backup task
ZPOOL MOUNT HOOK: 104 pre-start
104 is starting, mounting zpool 'backup'.
cannot export 'backup': pool is busy
cannot import 'backup': a pool with that name already exists
use the form 'zpool import <pool | id> <newpool>' to give it a new name
zpool import failed with error: 256
ERROR: Backup of VM 104 failed - hookscript error for 104 on pre-start: command '/var/lib/vz/snippets/backup-hook.pl 104 pre-start' failed: exit code 1
INFO: Failed at 2021-04-06 22:13:47

Some background on this particular job - I ran Proxmox Backup Server in a VM, then decided to try it in LXC. The snippet imports the backup pool on pre-start, then exports it on post-stop (allowing for hot-swap of backup drives). The snippet fails while backing up the deprecated, stopped VM in this case because the pool cannot be exported/imported since the new PBS container is running with a bind mount to it.

 

[oguz]

hi,

How do I distinguish between pre-start due to backup vs actually starting the VM/container?

with your hookscript like this:

#!/usr/bin/perl

use strict;
use warnings;

use PVE::QemuConfig;

print "GUEST HOOK: " . join(' ', @ARGV). "\n";

my $vmid = shift;

my $conf = PVE::QemuConfig->load_config($vmid);

my $phase = shift;

if ($phase eq 'pre-start') {
    print "doing preparations for $vmid\n";
    if ($conf->{lock} eq 'backup') {
            print "this is a backup!\n";
    } else {
            print "this isn't a backup, normal start!\n"
    }
}

 

[Elliott Partridge]

hi,



with your hookscript like this:

#!/usr/bin/perl

use strict;
use warnings;

use PVE::QemuConfig;

print "GUEST HOOK: " . join(' ', @ARGV). "\n";

my $vmid = shift;

my $conf = PVE::QemuConfig->load_config($vmid);

my $phase = shift;

if ($phase eq 'pre-start') {
    print "doing preparations for $vmid\n";
    if ($conf->{lock} eq 'backup') {
            print "this is a backup!\n";
    } else {
            print "this isn't a backup, normal start!\n"
    }
}

Thanks for that, works great! How would I adapt the script to handle both LXC and VMs? I have used this script on both. I see the equivalent perl module under /usr/share/perl5/PVE/LXC/Config.pm, but not sure how to check which one to use based on $vmid.

 

[Elliott Partridge]

Nevermind, I found an example of how to determine VM/LXC type in API2/BackupInfo.pm:

my $vmlist = PVE::Cluster::get_vmlist();
my $type = $vmlist->{ids}->{$vmid}->{type};
my $node = $vmlist->{ids}->{$vmid}->{node};

my $conf;
my $name = "";

if ($type eq 'qemu') {
    $conf = PVE::QemuConfig->load_config($vmid, $node);
    $name = $conf->{name};
} elsif ($type eq 'lxc') {
    $conf = PVE::LXC::Config->load_config($vmid, $node);
    $name = $conf->{hostname};
} else {
    die "VMID $vmid is neither Qemu nor LXC guest\n";
}