Linux filesystems

[anodos]

Revisiting this thread,
I made tests in latest Scale Nightly and left a comment in Jira ticket, which is otherwise is pretty stale:

Can you PM me a debug / logs of the samba crash? It's not surprising that this doesn't work because we make lots of assumptions that the underlying filesystem is ZFS, but crash is a bit excessive unless it's an assert because the FS isn't ZFS.

 

[InQuize]

Can you PM me a debug / logs of the samba crash? It's not surprising that this doesn't work because we make lots of assumptions that the underlying filesystem is ZFS, but crash is a bit excessive unless it's an assert because the FS isn't ZFS.

I uploaded debug in another ticket:

https://jira.ixsystems.com/browse/NAS-111095

I believe it is captured there, please check.

Show : smbd.log

[2021/06/18 01:52:14.132554, 0] ../../source3/lib/dumpcore.c:317(dump_core)
coredump is handled by helper binary specified at /proc/sys/kernel/core_pattern
'/mnt/twin/ext4-mount': not a ZFS filesystem
[2021/06/18 01:52:14.299565, 0] ../../source3/modules/smb_libzfs.c:233(get_zhandle)
get_zhandle: Failed to obtain zhandle on path: (/mnt/twin/ext4-mount)
[2021/06/18 01:52:14.299617, 0] ../../source3/modules/smb_libzfs.c:256(get_smbzhandle)
get_smbzhandle: Failed to obtain zhandle on path: [/mnt/twin/ext4-mount]: Invalid or incomplete multibyte or wide character
[2021/06/18 01:52:14.299661, 0] ../../source3/modules/smb_libzfs.c:1422(conn_zfs_init)
conn_zfs_init: Failed to obtain zhandle on connectpath: Invalid or incomplete multibyte or wide character
[2021/06/18 01:52:14.299706, 0] ../../lib/util/fault.c:172(smb_panic_log)

 

[anodos]

I uploaded debug in another ticket:

https://jira.ixsystems.com/browse/NAS-111095

I believe it is captured there, please check.

Show : smbd.log

[2021/06/18 01:52:14.132554, 0] ../../source3/lib/dumpcore.c:317(dump_core)
coredump is handled by helper binary specified at /proc/sys/kernel/core_pattern
'/mnt/twin/ext4-mount': not a ZFS filesystem
[2021/06/18 01:52:14.299565, 0] ../../source3/modules/smb_libzfs.c:233(get_zhandle)
get_zhandle: Failed to obtain zhandle on path: (/mnt/twin/ext4-mount)
[2021/06/18 01:52:14.299617, 0] ../../source3/modules/smb_libzfs.c:256(get_smbzhandle)
get_smbzhandle: Failed to obtain zhandle on path: [/mnt/twin/ext4-mount]: Invalid or incomplete multibyte or wide character
[2021/06/18 01:52:14.299661, 0] ../../source3/modules/smb_libzfs.c:1422(conn_zfs_init)
conn_zfs_init: Failed to obtain zhandle on connectpath: Invalid or incomplete multibyte or wide character
[2021/06/18 01:52:14.299706, 0] ../../lib/util/fault.c:172(smb_panic_log)

You were mounting an ext4 filesystem on top of your zpool?

 

[InQuize]

You were mounting an ext4 filesystem on top of your zpool?

Yes, as I described in the comment, /mnt/<pool>/ext4-mount, where 'twin' is zfs.
I was curious how NFS and SMB would react. As predicted, NFS does not give a damn and just works. I believe SMB could too, but it revealed those assumptions you mentioned. After all it's a throw away system and GUI specifically wants it that way in the current state of failsafe checks, so I simply experimented with tricking it. Obviously, I would not do so in production.

 

[anodos]

SMB and related protocols support exposing / taking snapshots and various other things. This means tighter integration with ZFS / libzfs in our case. Different filesystems also have different capabilities. I'll investigate and fix the issue, but leaving an SMB share with explicit ZFS configuration options in place when it's on a different filesystem is not really a recipe for success :)

 

[InQuize]

In case of ext3/ext4 filesystems it would be a win just to have limited basic file operations functionality. As I stated before, the main purpose of this use is to have cost-effective network attached file dump in the same system, not a competing fully fledged filesystem.
Would be nice if we could at least manually disable zfs specific functionality for SMB as we can tweak modules in CORE, if it won't ever be modified to recognize underlaying FS automatically.