VERITAS CLUSTER SERVER Resource Management - Veritas Cluster Server
This is a discussion on VERITAS CLUSTER SERVER Resource Management - Veritas Cluster Server ; Is there any tuning / enhancement documentation or notes regarding the VCS
and VxFS products. We are looking at any maens of tunning to reduce the amount
of system resources (CPU, Memory) the abovementioned products utilise on
our cluster environment. ...
-
VERITAS CLUSTER SERVER Resource Management
Is there any tuning / enhancement documentation or notes regarding the VCS
and VxFS products. We are looking at any maens of tunning to reduce the amount
of system resources (CPU, Memory) the abovementioned products utilise on
our cluster environment. We are running the products on two HP rp8400s using
HP-UX 11i, Orcale 9i. There are 102 Oracle 9i databases running split in
half on the two servers.
Are the following the areas we should be looking at:
• Default Parameters?
• Default Polling :VCS, VxFS
• Failover Resource polling
• LUN monitoring
• Database monitoring ( Oracle Agent))
-
Re: VERITAS CLUSTER SERVER Resource Management
Each resource type (mount or disk or ...) has a monitor interval. This
is "how often" the resource will be monitored (the resource agent spawns
a thread to do the monitor which might be either part of the agent
itself - a binary, or a Perl script). This is tunable (please see the
doco on VCS).
Another thing which you can tune as well, is the priority of the agents
(and thus resulting spwaned threads). Please see the doco on agent
process priority.
You can even go further (VCS 3.5 and above) by specifying the max.
number of threads each agent is allowed to spawn.
So, all-and-all a lot you can set for each agent.
(defaults can be seen by doing "hatype -display")
All the agents run as the "root" user, but some (like the Oracle one)
will do certain things (like deep monitoring) as the user/owner
specified in the VCS config.
If you want to, you can have a look at how the different resources are
actually monitored . If you're willing , you can make changes to the
monitor script, but beware that any change to the monitor scripts will
cause immediate invalidation of your support.
The monitor scripts are in /opt/VRTSvcs/bin/
where is something like Mount or Oracle or IP or .... (all the
different types of resources you get)
All I can say is that running 102/2 = 51 oracle instances on a box is
not a good thing (even if VCS is not in the picture). How about adding
more machines (domains/partitions) ?
Victor Mumbi wrote:
> Is there any tuning / enhancement documentation or notes regarding the VCS
> and VxFS products. We are looking at any maens of tunning to reduce the amount
> of system resources (CPU, Memory) the abovementioned products utilise on
> our cluster environment. We are running the products on two HP rp8400s using
> HP-UX 11i, Orcale 9i. There are 102 Oracle 9i databases running split in
> half on the two servers.
> Are the following the areas we should be looking at:
>
> • Default Parameters?
> • Default Polling :VCS, VxFS
> • Failover Resource polling
> • LUN monitoring
> • Database monitoring ( Oracle Agent))
>