[Linux-HA] Colocation confusion

[Christian Rishøj]

> It would be easier to give you an answer if you provide the
> CIBs as attachment - specially in the state where you do not
> understand why something happens ;-)

I am attaching the CIB with the status section stripped. It will now
be revealed that I actually have a few more resources configured,
which I in the interest of unclutteredness left out of the initial
presentation of the question.


> Here a couple things what comes me in mind.
>
> * i have no clue if colocation rules work on Cloned resources as you
> intended. The question is: if a clone primitive failes on one node,
> what impact does it have on the score for the clone ("database"
> or "webserver")?

I am *assuming* clones works with colocation, because I can do:

   % crm_resource -W -r webserver
   resource webserver is running on: ska
   resource webserver is running on: dub

and:

   % crm_resource -W -r database
   resource database is running on: ska
   resource database is running on: dub

Also, it seems that the WebIP isn't started on a server until the
database and webserver comes up, which would be a result of the
colocation constraints.


> * there is a nice tool called ptest from which you can get a lot of
> info (toke me quite a while until i was able to work with it).
>
> Here some filters i do when i debug a cluster:
> /usr/lib/heartbeat/ptest -L -VVVVVVVV 2>&1|grep test_expression|less
> /usr/lib/heartbeat/ptest -L -VVVVVVVV 2>&1|grep native_rsc_location|less
> /usr/lib/heartbeat/ptest -L -VVVVVVVV 2>&1|grep resource_node_score|less

Useful. I will apply these when a resource fails and look at the scores.


> * there are in general two ways to work with colocation:
> a) resources can not run anywhere except i tell them thay may run on a certain
> node
> b) resource can run evereywhere except i tell them to not run there.
> (i forgot the CIB attruibute which manages this behaviour but if you read the
> DTD it should be explained there)

I assume this is controlled by either specifying score="INFINITY" or
score="-INFINITY" in the rsc_colocation.


> ---> this has an impact when your depeneding resource fail over

This is what I was hoping. I hope to gain a little more insight later
today when I can run some tests.


Your help is greatly appreciated. Thank you.

Best,
Christian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cib.xml
Type: text/xml
Size: 9347 bytes
Desc: not available
Url : http://lists.community.tummy.com/pipermail/linux-ha/attachments/20070615/e7bba660/cib-0001.bin