pvestatd runs "zfs mount -a" over and over again -- how stopping it?

[hannes.sommer]

Background
It seems our `pvestatd` (from pve-manager pacakge, version 7.4-3) is calling `zfs mount -a` (about every 10 seconds) on one of our servers.

I come to this believe by running the following command repeatedly and observing the changing pids (here with a single example output):

$ pstree -pla |grep -B1 zfs ... -- |-pvestatd,10546 | `-zfs,1022591 mount -a ...

Unfortunately it has so many zfs filesystems (thanks to docker using it as backing storage) that a single invocation can take almost 10 seconds while eating 100% of a core. As a consequence we are basically losing one core to this rather useless action (all ZFSs are mounted anyways).

Questions
1. Can somebody help me disable this somehow?
2. Can somebody explain why it is running `zfs mount -a` in the first place? It seems very odd to me considering that `pvestatd` should be about collecting data and not altering the system.

 

[olidal]

Same issue here with pve 7.3-1, mount -a is using 100% of a core (72TB server, 957 datasets)...

 

[fiona]

Hi,
the command should only be executed when a storage is not detected as mounted by a certain heuristic. Please share the output of

pveversion -v
pvesm status
cat /etc/pve/storage.cfg
grep <dataset> /proc/mounts

with <dataset> being the pool property from the storage configuration, e.g. for the following configuration

zfspool: new
    pool myzpool/new
    content images,rootdir
    mountpoint /myzpool/new
    nodes pve701,pve702
    sparse 0

the command would be grep myzpool/new /proc/mounts.