Container filesystem corrupting after every shutdown (with SMB shared datasets, but not NFS)

[Jibak]

Brand new install of Proxmox on a R720 PowerEdge, everything up-to-date.

2 x SSD as mirror:
- Proxmox
- TrueNAS VM

4 x SSD as RAIDZ2:
- containers
- VMs

In TrueNAS, I passed through the 4 SSDs (cannot HBA/LSI passthrough because I only have the one controller). Created a RAIDZ2 pool and some datasets.

Shared the datasets with SMB, mounted them to Proxmox (via Storage->Add->SMB/CIFS). Installed containers (Ubuntu) on the dataset, which boot up fine. However, if I shutdown the containers and try to start them again, I get the following:

Aug 29 13:20:40 pve pvedaemon[25712]: starting CT 102: UPID:pve:00006470:0002106A:64ED6408:vzstart:102:root@pam:
Aug 29 13:20:40 pve pvedaemon[2472]: <root@pam> starting task UPID:pve:00006470:0002106A:64ED6408:vzstart:102:root@pam:
Aug 29 13:20:41 pve systemd[1]: Started pve-container@102.service - PVE LXC Container: 102.
Aug 29 13:20:41 pve kernel: loop1: detected capacity change from 0 to 25165824
Aug 29 13:20:41 pve kernel: EXT4-fs warning (device loop1): ext4_multi_mount_protect:326: MMP interval 42 higher than expected, please wait.

Aug 29 13:21:26 pve kernel: I/O error, dev loop1, sector 0 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 2
Aug 29 13:21:26 pve kernel: Buffer I/O error on dev loop1, logical block 0, lost sync page write
Aug 29 13:21:26 pve kernel: EXT4-fs (loop1): I/O error while writing superblock
Aug 29 13:21:26 pve kernel: EXT4-fs (loop1): mount failed
Aug 29 13:21:26 pve pvedaemon[25712]: startup for container '102' failed
Aug 29 13:21:26 pve pvestatd[2439]: unable to get PID for CT 102 (not running?)
Aug 29 13:21:26 pve pvedaemon[2472]: <root@pam> end task UPID:pve:00006470:0002106A:64ED6408:vzstart:102:root@pam: startup for container '102' failed
Aug 29 13:21:26 pve pvestatd[2439]: status update time (45.289 seconds)
Aug 29 13:21:27 pve pmxcfs[2302]: [status] notice: RRDC update error /var/lib/rrdcached/db/pve2-storage/pve/local: -1
Aug 29 13:21:27 pve pmxcfs[2302]: [status] notice: RRDC update error /var/lib/rrdcached/db/pve2-storage/pve/local-zfs: -1
Aug 29 13:21:27 pve pmxcfs[2302]: [status] notice: RRDC update error /var/lib/rrdcached/db/pve2-storage/pve/media: -1
Aug 29 13:21:27 pve pmxcfs[2302]: [status] notice: RRDC update error /var/lib/rrdcached/db/pve2-storage/pve/sucks: -1
Aug 29 13:21:27 pve systemd[1]: pve-container@102.service: Main process exited, code=exited, status=1/FAILURE
Aug 29 13:21:27 pve systemd[1]: pve-container@102.service: Failed with result 'exit-code'.
Aug 29 13:21:27 pve systemd[1]: pve-container@102.service: Consumed 1.463s CPU time.
Aug 29 13:21:31 pve kernel: CIFS: VFS: No writable handle in writepages rc=-9

If I run

pct fsck 102

they can boot up correctly. Although upon shutdown, the problem returns and the above error appears on startup. This is limited to containers, as VMs start/shutdown with no issues.

As a test I shared the same datasets with NFS, installed containers, and the problem does not appear (they can start/shutdown). Issue only appears when shared via SMB.

I would stick with NFS as a work-around, however I'd like to connect my Windows 10 Home PC to the shares (which requires SMB). Have spent many days on this, so am reaching out for help. Thanks!

 

[floh8]

think different! with zfs u can create multi datasets. so use one with nfs for vms and create a second one for smb! Its such simple.
But better is: i would advice not to use truenas. use the zfs function of proxmox. with local zfs u have additionally snapshots with containers and no additional layer with nfs and qcow2. u are faster.

 

[Jibak]

think different! with zfs u can create multi datasets. so use one with nfs for vms and create a second one for smb! Its such simple.
But better is: i would advice not to use truenas. use the zfs function of proxmox. with local zfs u have additionally snapshots with containers and no additional layer with nfs and qcow2. u are faster.

Thanks for the reply. I ended up going a simpler approach yes. One dataset shared via NFS store VMs/LXCs, another dataset shared via SMB for file sharing/storing. Makes more sense in hindsight!

I am enjoying TrueNAS because having the GUI (for me at least) makes setup/sharing/management of pools and datasets much easier. Speed is not a major concern for me, and I haven't noticed any issues as yet anywhere.