Hi,
On Mon, Jul 28, 2008 at 05:08:11PM +0200, Bj?rn Boschman wrote:
> Hi,
>
> I'm trying to build and test the following HA scenario:
>
> * 2 nodes using crm
>
> Using the following resources:
>
> * DRBD8 as master/slave
> * NFS server depending on DRBD
>
> * Mysql master/master replication (only ip-failover)
>
> * OpenSER active/hotstandby (only ip-failover)
>
> * Maybe some more active/active with ip-failover services later on
>
> Therefor I created the following resources and constraints:
>
> http://prod01.nfon.net/~nfon/resources.xml
> http://prod01.nfon.net/~nfon/constraints.xml
Looks mostly OK to me. Did you check the drbd howto with v2 at
linux-ha.org? You should increase timeouts, in particular those
which are lower than 10s. Also, does the openser LSB script
support the "promote" action? The first thing you should do is
check if all your resources can start, stop, and monitor or
status. Test in particular thoroughly the LSB class resources.
> For testing purposes all resources are created with meta-attribute
> target_role=stopped.
>
> SW versions:
> * heartbeat 2.1.3
> * pacemaker 0.6.5 (although there are several conflicts with existing
> heartbeat installation (duplicate files))
That shouldn't happen. If you're installing pacemaker then
download the heartbeat packages from the same place where you got
the pacemaker packages.
> So now my questions:
>
> 1.
> Would this in general make sense (also subjective meanings are welcome)
If it does what you need :)
> 2.
> When I start a resource with existing rsc_order dependency
> (e.g. mysql-ip) - should it start the dependend resources
> (override target_role)?
I don't think so. At any rate, it doesn't make sense to start a
resource and not all its dependencies.
> Currently I have to set
> target_role=started for both resources.
Yes.
> 3.
> I wanted to test that configuration using CTS but I got several
> errors.
Which errors?
> Maybe someone can give me a hint on CTS test options,
> or some things to keep in mind (beside global syslogging which
> is already working)
This is how I run CTS with a custom configuration:
nohup python /usr/share/pacemaker/cts/CTSlab.py -v2 \
--facility local7 --logfile /var/log/ha-log \
--stonith yes --fencing yes > ctsrun.out 2>&1 &
stonith/fencing is optional, but strongly recommended. Specify your
stonith device and options using
--stonith-type type
--stonith-args name=value
There's usage in /usr/share/pacemaker/cts/CTSlab.py, but it's
for whatever reason hard to get at. The CTS has been mainly used
by developers, so it's not exactly friendly.
Thanks,
Dejan
> If you need more input I can always put it online.
>
> Thanks for any replies!
>
>
> Mit freundlichen Gr??en
>
> Bj?rn Boschman
>
> nfon AG
> Nymphenburger Str. 70
> D-80335 M?nchen
>
> fon +49 (0)89 453 00-145
> fax +49 (0)89 453 00-100
> mail [EMAIL PROTECTED]
> web www.nfon.de
>
> Vorsitzender des Aufsichtsrats: Prof. Dr. Henning Werner
> Vorst?nde: Fabian Hoppe, Mathias Edelmann, Marcus Otto
>
> Sitz der Gesellschaft M?nchen
> Amtsgericht M?nchen, HRB 168022
> USt-ID DE254495743
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
