Yep, you do.
- The DAS overhead will more a RAM overhead issue than a CPU overhead
issue (IMHO).
- However, the larger concern will be when appserver instances become
very busy handling user load and consuming most of the server
resources, potentially interfering with your ability to use the DAS to
manage the cluster, deploy apps, etc. Appropriate capacity planning
and/or resource management capabilities in some Operating Systems
(such as Solaris) can address this issue.
- In addition, what is your recover scenario if the server with the
DAS (and a cluster node) fails? IMHO, a best practice for a mission-
critical production deployment is to:
1) Put the DAS on a separate host (can be shared with other
applications, preferably not cluster instances).
2) Back up the DAS configuration regularly
3) Have a DAS recovery scenario in place in event of a hardware
failure. It may be a restore from tape. It may be exporting the DAS
configuration regularly and importing it to a clean install. I
believe the High Availability Admin Guide and Deployment guide can
help here. Note, a downed DAS does not take down the running cluster,
but it does take away the ability to manage the cluster until a DAS
is back up and running.
On Feb 20, 2008, at 3:00 AM, glassfish_at_javadesktop.org wrote:
> Thank you...
>
> Now i have DAS on the first node of the cluster, and all works fine
> (same subnet...)
>
> But, do i have to think about an overhead of perfomrance on the
> cluster having DAS on a node instead of a standalone machine?
> [Message sent by forum member 'shambola' (shambola)]
>
> http://forums.java.net/jive/thread.jspa?messageID=259861
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: users-help_at_glassfish.dev.java.net
>