[Samba] MacOS clients - best options

[Tobias Hachmer]

Hi Thomas, hi Ralph,

thanks for pointing to the CephFS issue.

The problem still exists here. Our CEPH Cluster runs Ceph version 17.2.6 
on Debian Bullseye with kernel 5.10.0-22-amd64. For samba we use version 
4.18.2 (sernet packages) also on Debian Bullseye, kernel 5.10.0-22-amd64.

We currently test the workaround with streams_depot, but have problems 
here, also.

We use the group sticky bit extensivly and rely on group permissions. 
The relevant samba mask config is:

---
	force create mode = 0660
	force directory mode = 2770
	create mask = 0664
	directory mask = 2775
---

But the streams_depot VFS module creates the .streams and subdirectories 
with g=r-s, so I assume the mac clients have a permission problem and 
cannot write the metadata. Copying Files to a samba share fails with 
problem "-50". Sometimes the files are still written, sometimes not. 
After adjust the permissions to g=rws for the .streams directory 
including subdirs, everything is fine. But this is not a solution 
because we have to adjust the permissions regularly for new created dirs.

Are the permissions configurable for the streams_depot module or do we 
have to adjust our masks?

Nevertheless, we are heavily interested to solve the root cause and are 
willing to contribute information to the cephfs issue tracker.

Tobias


Am 5/15/23 um 12:00 schrieb Thomas Hukkelberg via samba:
> Thank you Ralph for pointing to the Ceph issue -- I'll followup the Ceph bugreport now. We run Ceph 16.2.11 on Linux 5.13.19/6.2.9 with Samba 4.17.6 and experience this issue several times a day. If we copy 50 small files, about 5% remains grayed out.
> 
> However, when validating this issue I did an interesting finding -- the brokMACS sequence seems only to prevail on file types that are known to macOS, like png, pdf, wav etc. When copying files that have unknown or no extension I never see the brokMACS sequence. Could the fact that Finder reads back the file to generate previews somehow mess up the clearing of the brokMACS sequence?
> 
> Tobias, do you see the same behavior?
> 
> --thomas
> 
> 
>> 11. mai 2023 kl. 22:12 skrev Ralph Boehme <slow at samba.org>:
>>
>> On 5/11/23 15:35, Thomas Hukkelberg via samba wrote:
>>> We have exact same problem but never really found the underlying
>>> issue and whether it's a cephfs bug, a samba bug or only related to
>>> vfs_fruit.
>>
>> it's a known subtle Ceph issue I'm afraid:
>>
>> https://tracker.ceph.com/issues/50719
>>
>> If the issue still exists with a newer Ceph version then mentioned in the bugreport, please post a followup in the bugreport so Jeff is made aware that the issue is still relevant and severe.
>>
>> -slow
>>
>> -- 
>> Ralph Boehme, Samba Team                 https://samba.org/
>> SerNet Samba Team Lead      https://sernet.de/en/team-samba
>> SAMBA+ Samba packages                   https://samba.plus/
>>
> 
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 2773 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/samba/attachments/20230516/b3cc7c56/smime.bin>