[DRBD-user] Use and usage of "call-pri-lost-after-sb" / How to outdate a peer

Kaloyan Kovachev kkovachev at varna.net
Thu Apr 28 11:05:19 CEST 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hi,

On Wed, 27 Apr 2011 19:18:39 +0000, "Stallmann, Andreas"
<AStallmann at CONET.DE> wrote:
> Good evening!
> 
> I'm new to this mailing list, so please be gentle. :) If I'm asking
> questions which are easily answered by a "RTFM" don't hesitate to point
me
> to the FM; I'm an eager reader. :)
> 
> We're using drbd in a two-node-cluster setup with pacemaker/corosync.
DRBD
> is used in a classic active/passive-fashion. Currently I'm running into
> trouble, when a node is disconnected (with both communication channels)
> from the cluster and comes back again (thus causing a split-brain).
Related
> to that, I've two questions:
> 
> 1. Which program is called, when policy for "after-sb-2pri" is set to
> "call-pri-lost-after-sb"?
> 
> The man for drbd.conf says "Call the "pri-lost-after-sb" helper program
on
> one of the machines. This program is expected to reboot the machine,
i.e.
> make it secondary. " Well... which program would that be? I  find the
> following in /usr/lib/drbd (OS is OpenSuSE 11.3):
> 
> cms-appl01:/usr/lib/drbd # ls
> crm-fence-peer.sh             notify-pri-lost.sh
> crm-unfence-peer.sh           notify-pri-on-incon-degr.sh
> notify-emergency-reboot.sh    notify-split-brain.sh
> notify-emergency-shutdown.sh  notify.sh
> notify-io-error.sh            outdate-peer.sh
> notify-out-of-sync.sh         snapshot-resync-target-lvm.sh
> notify-pri-lost-after-sb.sh   unsnapshot-resync-target-lvm.sh
> 
> Which of those would I have to define in the "handlers" section (if any
of
> those) for "pri-lost-after-sb" to reboot the machine? If there is no
such
> script in the distribution, has any of you written such a script and
could
> provide me with it? (Sorry for my bad English, but I'm getting awfully
> tired.)
> 
> 2. At some point I found in my logs, that the data of one of the peers'
> data is "invalid" and that the cluster is not synced, before the data is
> outdated. In earlier distributions of drbd, there was a "dopd" (drbd
peer
> outdater daemon). This seems not to exist anymore. Is the
outdate-peer.sh
> script a replacement for dopd? And, if yes, which parameter in which
> section of the drbd.conf makes drbd use it?
> 

Take a look in the example config - in the section handlers:
pri-lost-after-sb "/usr/lib/drbd/notify-pri-lost-after-sb.sh;
/usr/lib/drbd/notify-emergency-reboot.sh; echo b > /proc/sysrq-trigger ;
reboot -f";


> TNX in advance,
> 
> Andreas
> 
> ------------------------
> CONET Solutions GmbH, Theodor-Heuss-Allee 19, 53773 Hennef.
> Registergericht/Registration Court: Amtsgericht Siegburg (HRB Nr. 9136)
> Gesch?ftsf?hrer/Managing Directors: J?rgen Zender (Sprecher/Chairman),
> Anke H?fer



More information about the drbd-user mailing list