[Gluster-users] [EXT] upgrading OS on gluster servers

[Strahil Nikolov]

On an rpm-based system the following can give you an overview which dirs are being used by gluster:

rpm -ql (rpm -qa | grep gluster)

Sadly I got no cluster nearby to check it myself.

Best Regards,
Strahil Nikolov






В събота, 15 май 2021 г., 10:57:40 ч. Гринуич+3, Stefan Solbrig <stefan.solbrig at ur.de> написа: 





Hi Strahli,

It's a distributed-only volume (no replication). So heal or restore from backup is not an option (it's a really big FS).  The question is rather, if /etc/gluster  and /var/lib/glusterd are the only directories that are relevant, or if there are other directories.   
(I know that other distributions [commerical] have a longer support, but still, eventually, one will have to upgrade the server os or repace a server.)

best wishes,
Stefan


-- 
Dr. Stefan Solbrig
Universität Regensburg, Fakultät für Physik,
93040 Regensburg, Germany
Tel +49-941-943-2097



> Am 13.05.2021 um 19:27 schrieb Strahil Nikolov <hunter86_bg at yahoo.com>:
> 
> Hi Stefan,
> 
> add-brick requires FS without any extended file attributes which means that you start "fresh". You have to first remove the brick (to shrink the volume), detach the peer and after reinstall add the peer and expand the volume via add-brick (don't forget to recreate the FS).
> 
> If you have a lot of data, the restore from backup approach should save you a lot of healing, but you need to test it on a test setup (just to be on the safe side).
> 
> You didn't mention the volume type.
> In distributed volumes, I would go with remove-brick (start/commit), detach, peer probe ,add-brick .
> For replica volumes, you can try the backup/restore procedure.
> 
> Best Regards,
> Strahil Nikolov
>  
> 
>>  
>>  
>> On Wed, May 12, 2021 at 13:03, Stefan Solbrig
>> <stefan.solbrig at ur.de> wrote:
>> 
>> 
>>  
>> Hi Strahli,
>> 
>> Thank you for the quick answer!  Sorry I have to ask again: as far as I can see, Gluster keeps all information about peers, bricks, in /var/lib/glusterd.  So if I migrate to a new OS, it seems that I have to restore them. Or would you suggest rather to re-generate them by repeating all "peer probe ..."  and "volume brick-add ..." commands?
>> 
>> best wishes,
>> Stefan
>> 
>>  
>> -- 
>> Dr. Stefan Solbrig
>> Universität Regensburg, Fakultät für Physik,
>> 93040 Regensburg, Germany
>> Tel +49-941-943-2097
>> 
>> 
>> 
>>> Am 11.05.2021 um 13:46 schrieb Strahil Nikolov <hunter86_bg at yahoo.com>:
>>> 
>>> Hi Stefan,
>>> 
>>> 
>>> I would backup Gluster's dir in /etc .
>>> You don't need to restore any configuration files after the update, but it's good to have them backed up.
>>> 
>>> Best Regards,
>>> Strahil Nikolov
>>> 
>>> 
>>> 
>>>>  
>>>>  
>>>> On Tue, May 11, 2021 at 10:11, Stefan Solbrig
>>>> <stefan.solbrig at ur.de> wrote:
>>>> 
>>>> 
>>>>  
>>>> Dear all,
>>>> 
>>>> I was wondering what is the prefered way to upgrade the server OS (not glusterd) for a GlusterFS. 
>>>> I'm running a distributed-only system (no replication) on centos 7, planning an upgrade to centos 8 stream.
>>>> 
>>>> I suppose a possible way is like this:
>>>> 
>>>> * unmount file system on all clients
>>>> * stop cluster
>>>> * copy data in /var/lib/glusterd
>>>> * upgrade all servers
>>>> * restore data in /var/lib/glusterd
>>>> 
>>>> Could you please advise me if: 
>>>> - the files in /var/lib/glusterd  are all that is needed?
>>>> - or can I regenerate these from other data?
>>>> 
>>>> best wishes,
>>>> Stefan
>>>> 
>>>> 
>>>> ________
>>>> 
>>>> 
>>>> 
>>>> Community Meeting Calendar:
>>>> 
>>>> Schedule -
>>>> Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC
>>>> Bridge: https://meet.google.com/cpu-eiue-hvk
>>>> Gluster-users mailing list
>>>> Gluster-users at gluster.org
>>>> https://lists.gluster.org/mailman/listinfo/gluster-users
>>>> 
>>>> 
>> 
>> 
>>