Custom HA agent - Red Hat Linux Cluster

Linux and Unix-Like Red Hat
1

Hi experts, I have some custom application which I need to make Highly Available using red hat cluster service. How do I do it? i know in /usr/share/cluster i shall find HA agents for well known services like Apache or Sybase but I want to write HA agent for my own. I tried looking up on internet, and red hat documentation but couldn't find any systematic resource for the same. In system-config-cluster; I see a resource named script. Do I have to use this script resource for my custom application? If yes, I already did that. I wrote a script using apache.sh (found in /usr/share/cluster) and provided path of that script in resource config ( custom script is wrapper to start httpd daemon which bind to loop back address,stop this daemon and check its status). But i am unable to start it through cluster, i have added some ocf_log messages in script(for debugging) and they are not coming in /var/log/messages either. Any help, links, pointers would be greatly appreciated

2

Does the attached configuration guide for Red Hat clusters help?

3

Hi Neo

Cluster Admin Guide doesn't speak about setting up 'custom HA agent' or HA for custom applications.

Referring to this guide, i have successfully set up my 2-node red hat cluster; and have set up Apache service for HA

Now I want to learn or know how to make custom HA scripts, there is apache.sh in /usr/share/cluster and I was hoping to write my own using it, i did it as well however it seem not working.

I need some guidance as to how to achieve this(writing custom HA script), any inputs will be of great help.

4

Have you looked at the templates provided by Red Hat for writing custom failover scripts?

5

Hi fpmurphy

I looked in /usr/share/cluster/script.sh (and I think that's the template for custom app failover?). Here is how it looks like

#
# Script to handle a non-OCF script (e.g. a normal init-script)
#

LC_ALL=C
LANG=C
PATH=/bin:/sbin:/usr/bin:/usr/sbin
export LC_ALL LANG PATH

. $(dirname $0)/ocf-shellfuncs

meta_data()
{
    cat <<EOT
<?xml version="1.0"?>
<resource-agent version="rgmanager 2.0" name="script">
    <version>1.0</version>
.....
}

case $1 in
        meta-data)
                meta_data
                exit 0
                ;;
        *)
                ;;
esac

[ -n "${OCF_RESKEY_file}" ] || exit $OCF_ERR_ARGS      # Invalid Argument
[ -f "${OCF_RESKEY_file}" ] || exit $OCF_ERR_INSTALLED # Program not installed
[ -x "${OCF_RESKEY_file}" ] || exit $OCF_ERR_GENERIC   # Generic error

# Don't need to catch return codes; this one will work.
ocf_log info "Executing ${OCF_RESKEY_file} $1"
${OCF_RESKEY_file} $1

declare -i rv=$?
if [ $rv -ne 0 ]; then
        ocf_log err "script:$OCF_RESKEY_name: $1 of $OCF_RESKEY_file failed (returned $rv)"
        exit $OCF_ERR_GENERIC
fi

Using above template i have created 'dummy_service.sh' and added status, start, stop, verify-all, monitor options in it (looking at apache.sh) to start, stop, monitor, verify my application is running (actually i am just trying to start httpd and stop it for now before going to more complex custom app)

If you like I can attach that script too.

Thanks

---------- Post updated 23-09-10 at 12:57 AM ---------- Previous update was 22-09-10 at 11:25 PM ----------

Alright it seems like 'script.sh' is not template for custom failover script

It is actually a resource agent for 'script' resource, in this script resource we need to provide path for /etc/init.d/<my script>

However this is a bit limiting, there are some unknown caveats like how to find out a node from which service is failed over (name of node on which service was previously running) and accessing shared resources (e.g. floating IP addresses)

---------- Post updated at 06:30 AM ---------- Previous update was at 12:57 AM ----------

I was able to set up custom HA agent; looking at few more options,

Thanks