NFSv4 cross-filesystem not working after upgrade to 13

[rockybulwinkle]

When I was on TrueNAS Core 12, I had gotten cross-filesystem mounts working with NFSv4.

What this means is, I could have my "ROOT" dataset, and a number of datasets underneath it, accessible by adding a single path in the TrueNAS NFS shares. For example, here are some of my datasets:

Code:

data0/ROOT
data0/ROOT/games
data0/ROOT/movies

Under TrueNAS 12, I was able to export a single ZFS dataset (data0/ROOT) and have access to all datasets underneath it through NFS. The client automounts the different ZFS datasets in both 12 and 13. Below I show what happens under 13, including the I/O error.

Code:

[root@saturn data0]# mount | grep data0
192.168.2.229:/mnt/data0/ROOT on /data0 type nfs4 (rw,relatime,vers=4.2,rsize=131072,wsize=131072,namlen=255,hard,proto=tcp,nconnect=16,timeo=600,retrans=2
,sec=sys,clientaddr=192.168.2.198,local_lock=none,addr=192.168.2.229)
[root@saturn data0]# ls /data0/games
ls: reading directory '/data0/games': Input/output error
[root@saturn data0]# mount | grep data0
192.168.2.229:/mnt/data0/ROOT on /data0 type nfs4 (rw,relatime,vers=4.2,rsize=131072,wsize=131072,namlen=255,hard,proto=tcp,nconnect=16,timeo=600,retrans=2
,sec=sys,clientaddr=192.168.2.198,local_lock=none,addr=192.168.2.229)
192.168.2.229:/mnt/data0/ROOT/games on /data0/games type nfs4 (rw,relatime,vers=4.2,rsize=131072,wsize=131072,namlen=255,hard,proto=tcp,nconnect=16,timeo=6
00,retrans=2,sec=sys,clientaddr=192.168.2.198,local_lock=none,addr=192.168.2.229)

For reference, here is my /etc/exports under TrueNAS 13:

Code:

V4: / -sec=sys
/mnt/data0/ROOT -alldirs -maproot="root":"wheel"

And in case it matters, my systemd unit file for mounting NFS on my client:

Code:

[Unit]
Description=Mount Share at boot
After=network-online.target

[Mount]
What=192.168.2.229:/mnt/data0/ROOT
Where=/data0
Options=nconnect=16
Type=nfs4
TimeoutSec=10

[Install]
WantedBy=multi-user.target

The funny thing is, if I create a new dataset under data0/ROOT through the command line, it is accessible through NFS as I expect, up until a reboot. After reboot, the dataset gives the same I/O error as before. So it's definitely still possible to get the old behavior.

As a workaround, I can add the ZFS datasets as additional paths to share in NFS, but I really need to get the old behavior working--at my work, we also use TrueNAS Core and keep many tens of paths and it would be cumbersome to add each one individually when we eventually upgrade to 13.

 

[BensonHo]

i also want to all exports can be mounted together in a dir tree structure ，can i just export the parent dataset on truenas ,
my client would ls all the sub-dataset and dirs. after client just mounted this parent dataset export path

nfsv4 support “ Export:All exports can be mounted together in a directory tree structure as part of a pseudo-filesystem “ ？
client should use mount -t nfs4 ？