[Linux-cluster] 2 node cluster w/o shared storage and custom script

Rafael Micó Miranda rmicmirregs at gmail.com
Mon Oct 19 21:54:11 UTC 2009


Hi all,

I've been requested to design a solution for a 2 node, active-passive,
cluster of Maven (http://maven.apache.org/)

The cluster will be formed by:

- 2 front-end nodes in which the Maven service will be run, in an
active-passive configuration (active-backup)
- I will use HP-ilo fencing devices as the nodes are HP servers
- a NFS mountpoint in which Maven can storage its files, served by a NAS
cluster
- a SQL database

The need of an active-passive configuration is because, as I've been
told, the front-end nodes cannot run Maven at the same time because it
will cause database corruption.

I'm accustomed to clusters in which nodes share a shared storage node,
so i could use a Qdisk daemon in a 2 node cluster in addition to fencing
to keep cluster quorate. This is not the situation as i will not have
available a shared storage device, and this will be the first cluster
using CMAN to be used in the company (i built the older ones with
Linux-HA).

I have some doubts about the better way to achieve this configuration,
and I will like to share it with you and get your recommendations:

1.- "Scrip resource": as official RHEL 5.4 documentation says: "The
script resource allows a standard LSB-compliant init script to be used
to start a clustered service."

I'm used to LSB scripts as part of the Linux-HA project:
http://www.linux-ha.org/LSBResourceAgent

Is there any further information of the LSB compliance for resource
scripts in RHEL ?


2.- Service operational status: does rgmanager implement any way to
check the status of the "script" resources? If so, will the "script"
resource need an "status|monitor" method? How often will rgmanager check
the status of the resource?


3.- NFS resource operational status: same as above: does rgmanager
implement any way to check the status of the "nfs mountpoint" resources?
How often will rgmanager check the status of the resource?


4.- Qdisk: As far as i know, i will not be able to use Qdisk here. I
plan to use the 2 node cluster configuration:

<cman expected_votes="1" two_node="1"/> 

Is there any other consideration to take into account to keep the
cluster quorate?



Any help/recommendations will be appreciated.

Cheers,

Rafael

-- 
Rafael Micó Miranda




More information about the Linux-cluster mailing list