[Linux-cluster] Can anyone help tackling issues with custom resource agent for RHCS?

[Ralph.Grothe at itdz-berlin.de]

Hello,

I have written a custom resource agent for Informix servers in
accordance to the OCF RA Developer's Guide 
http://www.linux-ha.org/doc/dev-guides/ra-dev-guide.html
but I have some doubt if it properly integrates with RHCS's
idiosyncrasies.

I would have liked to save me the work finding out all the RedHat
deviations, and would have taken a readily supplied agent for
Informix Server resources if only I had found one.

I found that I not only had to write the agent and simply place
it in /usr/share/cluster 
(as far as I understand this would have sufficed in a Pacemaker
cluster, albeit under a different directory path) 
I also had to tinker with
/usr/share/system-config-cluster/misc/cluster.ng in order to have
my extensions pass the XML validity check.

Now my new <ifxdb/> tag is accepted in cluster.conf along with
its paramtrezation via in the agent's meta-data dump as well as
my definitions in cluster.ng,
and I was able to successfully commit an ccs_tool update.


These are the attributes and actions my agent supports so far:

# /usr/share/cluster/ifxdb.sh meta-data|grep -E
'(parameter|action) name'
        <parameter name="server" primary="1" unique="1">
        <parameter name="home" required="1">
        <parameter name="config" required="1">
        <parameter name="version">
        <parameter name="locale">
        <parameter name="user">
        <parameter name="dbdate">
        <parameter name="dbmoney">
        <parameter name="log">
        <action name="start" timeout="60"/>
        <action name="stop" timeout="60"/>
        <action name="recover" timeout="180"/>
        <action name="status" depth="1" timeout="10"
interval="15"/>
        <action name="monitor" depth="10" timeout="20"
interval="30"/>
        <action name="meta-data" timeout="5"/>
        <action name="verify-all" timeout="5"/>






rg_test shows its rules and I am able to start, stop and monitor
(or status in RHCS mannerism) my Informix resources
like e.g.

# rg_test test /etc/cluster/cluster.conf start ifxdb ju_09tcp


I can also enable, stop and relocate every service that contains
an instance of my ifxdb resource via clusvcadm.


As I made ample use of the ocf_log function 
(I prefer to have some meaningful output from the agents in the
logs)
at the command line when invoked through rg_test the agent
verbosely reports what it's doing.

What puzzles me is that I cannot see any of these logging entries
when my agent is run by clurgmgrd.
That it is run at all I can verify as on e.g. service startup the
Informix server instance is successfully started.

What puzzles me even more is that I cannot see any entries in
/var/log/messages from the regular monitoring invocatins of the
agent at given intervals according to above meta-data dump, so as
if no resource monitoring checks are at all performed.

And even worse, when I shutdown an Informix instance manually
(e.g. omnode -ky) it doesn't get restarted by clurgmgrd even
though *no* __independent_subtree attribute is defined in the
ifxdb tags of cluster.conf.

So although I can manually start, stop and relocate the affected
services through clusvcadm, the whole rgmanager HA treatment
seems dysfunctional to me.

Is there anything or anywhere that I have forgotten to also
manipulate to fully enable my custom resource agent?



Regards,
Ralph










--
Fon  +49 30 90222 6481
Fax  +49 30 90222 3151

IT-Dienstleistungszentrum Berlin
Anstalt öffentlichen Rechts
PB22 Unix/Linux/vSphere Administration
Berliner Str. 112-115
D-10713 Berlin
Handelsregisternr.: HRA 36349 B
Registergericht: Amtsgericht Charlottenburg
www.itdz-berlin.de
www.itdz.verwalt-berlin.de

Das ITDZ Berlin ist im Rahmen des "audit berufundfamilie" als
familienbewusstes Unternehmen zertifiziert.

Bitte prüfen Sie aus Gründen des Umweltschutzes, ob der Ausdruck
dieser E-Mail erforderlich ist.