Mailing List Archive

DRBD user guide - GFS2 - recommendation to lose data?
Section 10.2:
"Also, it is recommended to use some of DRBD’s features for automatic
recovery from split brain. To do all this, include the following lines
in the resource configuration:
...
By configuring auto-recovery policies, you are configuring effectively
configuring automatic data-loss! Be sure you understand the
implications."

If these options incur a risk of data loss, why are they recommended?
I'm wary of recommending that to our users.

https://linbit.com/drbd-user-guide/drbd-guide-9_0-en/#ch-gfs

--
Regards,

Reid Wahl (He/Him)
Senior Software Engineer, Red Hat
RHEL High Availability - Pacemaker

_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user
Re: DRBD user guide - GFS2 - recommendation to lose data? [ In reply to ]
On 8/2/22 22:48, Reid Wahl wrote:
> Section 10.2:
> "Also, it is recommended to use some of DRBD’s features for automatic
> recovery from split brain. To do all this, include the following lines
> in the resource configuration:
> ...
> By configuring auto-recovery policies, you are configuring effectively
> configuring automatic data-loss! Be sure you understand the
> implications."
>
> If these options incur a risk of data loss, why are they recommended?
> I'm wary of recommending that to our users.
>
> https://linbit.com/drbd-user-guide/drbd-guide-9_0-en/#ch-gfs

Hello again :)

As you've noticed, the "GFS" section of the DRBD User Guide could use
some attention. I believe it was written prior to GFS2's release (so
RHEL 5?).

Issues with prose aside, this is another case where clarification is
needed. In a properly configured Pacemaker + DRBD cluster (i.e. with
fencing) when the replication network splits DRBD will call for
Pacemaker to fence the peer. Both peers will have been Primary but
disconnected and therefore DRBD will be in a split-brain state when
the peer reboots and reconnects. It will reconnect in the Secondary
state and because of the `after-sb-1pri discard-secondary;` setting
the data on the returning Secondary node will be overwritten with the
data from the Primary. Technically, data on the returning peer could
be lost but its probably not anything you'd want to keep.

I believe the comment is alluding to more dangerous settings, such as
`after-sb-0pri discard-least-changes;`. For example, that setting in
an improperly configured cluster could result in a video of an
intern's cat (large) being saved while the CFO's pin number (small) is
discarded (or something equally dramatic).

I've opened an internal issue to clean up this section for future readers.

Best Regards,
Matt

--
Matt Kereczman – Solutions Architect
matt@linbit.com

LINBIT | Keeping the Digital World Running
HA - DR - SDS

LINBIT® is a registered trademark of LINBIT, Austria.
_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user
Re: DRBD user guide - GFS2 - recommendation to lose data? [ In reply to ]
On 8/16/22 16:58, Matt Kereczman wrote:
> On 8/2/22 22:48, Reid Wahl wrote:
>> Section 10.2:
>> "Also, it is recommended to use some of DRBD’s features for automatic
>> recovery from split brain. To do all this, include the following lines
>> in the resource configuration:
>> ...
>> By configuring auto-recovery policies, you are configuring effectively
>> configuring automatic data-loss! Be sure you understand the
>> implications."
>>
>> If these options incur a risk of data loss, why are they recommended?
>> I'm wary of recommending that to our users.
>>
>> https://linbit.com/drbd-user-guide/drbd-guide-9_0-en/#ch-gfs
<snip>
>
> I've opened an internal issue to clean up this section for future
> readers.

Wanted to let you know we pushed some adjustments to the language in
the GFS section of the DRBD user guide. Thanks for the feedback. If
you have any further recommendations don't hesitate to reach out.

--
Matt Kereczman – Solutions Architect
matt@linbit.com

LINBIT | Keeping the Digital World Running
HA - DR - SDS

LINBIT® is a registered trademark of LINBIT, Austria.
_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user
Re: DRBD user guide - GFS2 - recommendation to lose data? [ In reply to ]
On Thu, Aug 18, 2022 at 11:15 AM Matt Kereczman <matt@linbit.com> wrote:
>
> On 8/16/22 16:58, Matt Kereczman wrote:
> > On 8/2/22 22:48, Reid Wahl wrote:
> >> Section 10.2:
> >> "Also, it is recommended to use some of DRBD’s features for automatic
> >> recovery from split brain. To do all this, include the following lines
> >> in the resource configuration:
> >> ...
> >> By configuring auto-recovery policies, you are configuring effectively
> >> configuring automatic data-loss! Be sure you understand the
> >> implications."
> >>
> >> If these options incur a risk of data loss, why are they recommended?
> >> I'm wary of recommending that to our users.
> >>
> >> https://linbit.com/drbd-user-guide/drbd-guide-9_0-en/#ch-gfs
> <snip>
> >
> > I've opened an internal issue to clean up this section for future
> > readers.
>
> Wanted to let you know we pushed some adjustments to the language in
> the GFS section of the DRBD user guide. Thanks for the feedback. If
> you have any further recommendations don't hesitate to reach out.

Thanks so much for the clarification and doc updates!

I 100% sympathize BTW. It's daunting to keep docs up to date,
especially where they involve other vendors'/projects' software.

As an FYI for if there's time to do an overhaul later, Red Hat stopped
supporting CLVMD + GFS2 as of RHEL 8. Now we use lvmlockd + GFS2. Here
are the relevant portions of the RHEL 9 product docs:
- https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/9/html-single/configuring_and_managing_high_availability_clusters/index#assembly_configuring-gfs2-in-a-cluster-configuring-and-managing-high-availability-clusters
- https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/9/html/configuring_gfs2_file_systems/index

>
> --
> Matt Kereczman – Solutions Architect
> matt@linbit.com
>
> LINBIT | Keeping the Digital World Running
> HA - DR - SDS
>
> LINBIT® is a registered trademark of LINBIT, Austria.
> _______________________________________________
> Star us on GITHUB: https://github.com/LINBIT
> drbd-user mailing list
> drbd-user@lists.linbit.com
> https://lists.linbit.com/mailman/listinfo/drbd-user



--
Regards,

Reid Wahl (He/Him)
Senior Software Engineer, Red Hat
RHEL High Availability - Pacemaker

_______________________________________________
Star us on GITHUB: https://github.com/LINBIT
drbd-user mailing list
drbd-user@lists.linbit.com
https://lists.linbit.com/mailman/listinfo/drbd-user