Resize API endpoint (PUT) - async?

[muhaha]

Hello,
I am creating VMs (qemu) in Proxmox with Ansible.
1. Clone(API endpoint - POST) -> cloud-init template
2. Config(API endpoint - POST) -> cloned VM
3. Resize(API endpoint - PUT) -> scsc0 -> size=+10G


Clone request will return task ID in response, so I can check if VM was cloned to continue.
Config will not return anything usefull, but It will always success.
Resize will sometimes return /usr/bin/qemu-img resize -f raw /data/pool1/images/101/vm-101-disk-0.raw 10737418240' failed: got timeout and it does not matter if I am using size=+1G, size=10G, size=+10G.


I am assuming that VM is not ready for Resize yet, but I am not sure, how can I check it to run another step (Resize)

If I create request /let say 5min after VM config/ it will always return success. So there is some kind of background task which will need to be completed to Resize disk.

Any ideas?

 

[fabian]

we should probably change resize_vm to be async and fork a worker. could you file an enhancement request at https://bugzilla.proxmox.com ?

 

[muhaha]

https://bugzilla.proxmox.com/show_bug.cgi?id=2315

 

[helldorado]

hello,

I encountered the same problem to deploy via terraform. The timeout of the qemu-img resize command is 10s and impossible to overload. So I made a local patch to volume_resize function.

sub volume_resize {
    my ($class, $scfg, $storeid, $volname, $size, $running) = @_;

    die "can't resize this image format\n" if $volname !~ m/\.(raw|qcow2)$/;

    return 1 if $running;

    my $path = $class->filesystem_path($scfg, $volname);

    my $format = ($class->parse_volname($volname))[6];

    my $cmd = ['/usr/bin/qemu-img', 'resize', '-f', $format, $path , $size];

    run_command($cmd, timeout => 10);

    return undef;
}

Cf => https://github.com/proxmox/pve-storage/blob/master/PVE/Storage/Plugin.pm#L668-L684



It would be great to overload the timeout from the API.

 

[gca]

Hi,

I stumbled across this problem too.

My "workaround" was to wait for the iowait value to go below 2 (top -d2 -bn1 | grep -o -E '[[:digit:]]+.[[:digit:]]+ wa,' | sed 's/ wa,//') before trying to resize the disk so that the qemu-img command last less than 10 seconds.

Any news on https://bugzilla.proxmox.com/show_bug.cgi?id=2315 ?