Extremely slow disk move from CIFS/SMD

[splix]

I've just reinstalled the Proxmox, but before it I copied some of the disk of the existing machines to a NAS. Now I'm trying to move the disk back, to the new system.

It seem that Proxmox is waiting for something before actual moving, but it's unclear what. For a small disk, which is ~20Gb, it waits for about 10 minutes before starting an actual data copying. I've tried it a few times, and each time I see no traffic on the NAS side (i.e, no network traffic, not disk IO, etc) for the first 5-10 minutes. Then it starts copying, though the speed is about 1/3 of what it supposed to be.

But for a larger disk it takes forever. I waited for about 12 hours max, and a few times waited for an hour or two, and it never started actual copying. No traffic or requests on the NAS side.

I restarted the machine few times, checked all the network configurations, can access NAS from the command line, list and copy files, all looks fine. Even reinstalled Proxmox from scratch. Nothing helps.

What should I check to find out the reason of that?

I see in `pvestatd` the following:

pvestatd[1133]: unable to activate storage 'nas' - directory '/mnt/pve/nas' does not exist or is unreachable
pvestatd[1133]: got timeout

Just to clarify, this error repeats every minute.

And

# ls -l /mnt/pve/
total 0
drwxr-xr-x 2 root root 0 Sep  7 21:55 nas

# ls -l /mnt/pve/nas
...TIMEOUT...

But once I stop the moving task, I'm getting no timeout. I.e. `ls` on that dir shows the files on the remote side if there is no copying task. I can see the files content, create and delete them. Also I can create VMs from ISO stored on that NAS. I.e., it's not auth or access rights problem.

What could be a possible reason?

 

[splix]

I did a few more tests.

I mounted the nad share directly, i.e., as `mount -t cifs ... /mnt/nas` to a directory, and added this directory as a Proxmox storage. To my understanding Proxmox should treat is as any other directory.

Even though I can access/read/write this dir from command line, and Proxmox sees Disks and ISOs stored there.

But once I try to mvoe the disk (`qm disk move 101 unused4 data`) it gets stuck with everything related to that share. `ls -l /mnt/nas` takes forvever, as well as `pvesm status`.

 

[splix]

Ah, actually, `pvesm status` gives a result after a long wait:

# pvesm status
got timeout
Name             Type     Status           Total            Used       Available        %
data          zfspool     active     14864491571      7476138082      7388353489   50.30%
local             dir   disabled               0               0               0      N/A
local-lvm     lvmthin     active      1793077248        18648003      1774429244    1.04%
nas              cifs   disabled               0               0               0      N/A
nas-dir           dir   inactive               0               0               0    0.00%

Note `got timeout` message.

Once I Ctrl+C on the command that moved the data it starts working fast and gives a result like that (which is correct):

# pvesm status
Name             Type     Status           Total            Used       Available        %
data          zfspool     active     14864491249      3216232034     11648259215   21.64%
local             dir   disabled               0               0               0      N/A
local-lvm     lvmthin     active      1793077248        18648003      1774429244    1.04%
nas              cifs   disabled               0               0               0      N/A
nas-dir           dir     active      8589934592      2545243456      6044691136   29.63%

Note that for the `data` storage it actually reserves a space required (it's 50% instead of 21%), but stuck with data copy.

 

[RolandK]

have a look here, maybe it's related

https://forum.proxmox.com/threads/weird-samba-performance-bug-issue.154258/

 

[splix]

Interesting. It seems to be my case too. Thank you

Simple `touch` of the file didn't help, though, but at least I know where to look