[DRBD-user] Primary/Secondary issue

[Jakov Sosic]

On Mon, 5 Oct 2009 20:46:47 -0400
"Paul Stewart" <paul at paulstewart.org> wrote:

> Anyways, these are very high load servers and we are migrating to a
> new platform in the coming weeks - thankfully.  The problem is that a
> fail over condition occurred last week and since then the primary
> server and the secondary server don't see each other.  This doesn't
> appear to be a split-brain scenario.
> 
> Secondary Server doesn't see the primary - looks like it's waiting
> for a connection..

> Primary Server doesn't see secondary so has gone to standalone state
>  0: cs:StandAlone st:Primary/Unknown ds:UpToDate/Outdated   r---

You can try:
primary # drbdadm connect all

And if that does not work, then you must invalidate all data on
secondary node... This will cause full resync, so you will have
performance impact. If you run this, run it on saturday, or whenever
you expect the lowest load on that cluster.

secondary # drbdadm disconnect all
secondary # drbdadm secondary all
secondary # drbdadm -- --discard-my-data connect all



> Between these two servers is a private point to point Ethernet link
> that has no problems passing traffic, so network communication
> doesn't appear to be the issue.

> I want the primary server to remain primary but am concerned the
> secondary server will not try to come online no matter what I've
> tried.   Being new to drbd I'm sure it's just me ;) 

This is a matter of heartbeat if the secondary will try to come online.
I use RHCS and not heartbeat, so I don't know what the heartbeat
migration script does, but it shouldn't try to migrate if the resync is
not over. Also, search for a autofailback option in /etc/ha.d/ - if
that's zero, then it won't switch back at all, unless new primary fails.

Hope this helps.



-- 
|    Jakov Sosic    |    ICQ: 28410271    |   PGP: 0x965CAE2D   |
=================================================================
| start fighting cancer -> http://www.worldcommunitygrid.org/   |