[Samba] CIFS client mounts meta ops 30 times slow than server

[Jeremy Allison]

On Fri, Mar 10, 2023 at 01:41:57AM +0800, Shenyue Chen wrote:
>   Hi Jeremy, thanks for the quick response!
>   Yes, BeeGFS is a clustered file system. It's just behaving like a regular
>   FS after mounted. 
>   We mount it on the SMB server and re-export the mounted folder to the
>   linux CIFS clients.
>   Here is a config file on the SMB server FYI. Most of the options we keep
>   default. 

So you're comparing the Linux cifsfs performance
going to a Samba share that maps to a local filesystem
with a Samba share that maps to a BeeGFS mounted filesystem yes ?

You might want to try upgrading to 4.17.6 or 4.18.0 if you're
feeling adventurous :-).

 From the 4.18.0 release notes:

NEW FEATURES/CHANGES
====================

SMB Server performance improvements
-----------------------------------

The security improvements in recent releases
(4.13, 4.14, 4.15, 4.16), mainly as protection against symlink races,
caused performance regressions for metadata heavy workloads.

While 4.17 already improved the situation quite a lot,
with 4.18 the locking overhead for contended path based operations
is reduced by an additional factor of ~ 3 compared to 4.17.
It means the throughput of open/close
operations reached the level of 4.12 again.

>   [global]
>
>   ## Browsing/Identification ###
>
>   # Change this to the workgroup/NT-domain name your Samba server will part
>   of
>      workgroup = WORKGROUP
>
>   # server string is the equivalent of the NT Description field
>      server string = %h server (Samba, Ubuntu)
>
>      clustering = yes
>      idmap config * : backend = autorid
>      idmap config * : range = 1000000-1999999
>      netbios name = BeeGFS *
>      vfs objects = fileid
>      fileid:algorithm = fsid
>   [beegfs]
>   comment = BeeGFS
>   path = /mnt/beegfs
>   browsable = yes
>   writable = yes
>   read only = no