BeeGFS upgrade questions

[Steffen Grunewald]

Hi,


in addition to the question asked before, I've got two short questions
regarding upgrade:


(1) I presume there's no way to downgrade servers that have been upgraded?
(I couldn't find an explanation why, but I suspect that on-disk data
representations have been changed irreversibly - is this documented
somewhere?)

(2) How long does it take to perform an upgrade from 6.19 to 7.2.x, to
get the on-disk data converted properly? (Will that happen in a single
sweep, or by file access? Is there a rule of thumb as in "x minutes
per million inodes/chunks"?)


Thanks once more,

Steffen


--
Steffen Grunewald, Cluster Administrator
Max Planck Institute for Gravitational Physics (Albert Einstein Institute)
Am Mühlenberg 1 * D-14476 Potsdam-Golm * Germany
~~~
Fon: +49-331-567 7274
Mail: steffen.grunewald(at)aei.mpg.de
~~~

[Peter Grossoehme]

> -----Original Message-----
> From: fhgfs...@googlegroups.com <fhgfs...@googlegroups.com> On
> Behalf Of Steffen Grunewald
> Sent: Thursday, October 7, 2021 10:32 AM
> To: BeeGFS user list <fhgfs...@googlegroups.com>
> Subject: [beegfs-user] BeeGFS upgrade questions
>
> Hi,

Hi Steffen,

> in addition to the question asked before, I've got two short questions regarding
> upgrade:
>
>
> (1) I presume there's no way to downgrade servers that have been upgraded?
> (I couldn't find an explanation why, but I suspect that on-disk data
> representations have been changed irreversibly - is this documented
> somewhere?)

As far as I know the on-disk format has been changed which prevents a downgrade to previous version as soon as new data has been written to the filesystem theoretically.
I expect similar behaviour for metadata format because storage pools has been introduced with release 7.0. As far as I remember correctly metadata format is updated during user modification of files.
Furthermore the mgmtd store will be upgraded to latest version but it can be recovered as soon as you allow registration of new servers / targets after re-init - just the user and group quota limits will get loss if you don't have any csv backup via beegfs-ctl.
Finally, I would not really recommend to try a downgrade - the beegfs-meta daemon always reads the root directory during first start and might do already some small conversion at this time.

> (2) How long does it take to perform an upgrade from 6.19 to 7.2.x, to get the
> on-disk data converted properly? (Will that happen in a single sweep, or by file
> access? Is there a rule of thumb as in "x minutes per million inodes/chunks"?)

=> https://doc.beegfs.io/7.2/advanced_topics/upgrade.html
You can directly upgrade from 6.19 to 7.2.X - so after upgrading the packages and restarting all services there is no big "conversion process" starting.
"Old" data remains in previous on-disk format and new data will be stored in new style. This means BeeGFS 7.2.X can read and modify existing data in the 6.x release series on-disk data format.

All statements without 100% guarantee - use at your own risk! 😉


Best regards,

Peter
--
Head of Engineering

Tel.: +49 3722 528-45
Fax: +49 3722 528-15
E-Mail: peter.gr...@megware.com

MEGWARE Computer
Vertrieb und Service GmbH
Nordstraße 19
09247 Chemnitz-Röhrsdorf
Germany
http://www.megware.com

Sitz der Gesellschaft: MEGWARE Computer Vertrieb und Service GmbH, Nordstraße 19, 09247 Chemnitz-Röhrsdorf
Geschäftsführer: Jürgen Gretzschel, Steffen Eckerscham, André Singer, Axel Auweter, Amtsgericht Chemnitz HRB 584

Headquarters: MEGWARE Computer GmbH Vertrieb und Service, Nordstraße 19,
09247 Chemnitz-Röhrsdorf
Board of Directors: Steffen Eckerscham, André Singer, Axel Auweter County Court: Chemnitz (Germany) HRB 584