SUMMARY: Sun Cluster scdpmd (Disk Path Monitoring) high system load and core dumps

Great list!!!!

With thanks to Martin Pre_laber who responded almost instantly - We're dealing
with Bug ID 6563949,
http://sunsolve.sun.com/search/document.do?assetkey=1-66-201254-1

Impact:
Systems running Solaris Cluster 3.2, with the patches listed in
the "Contributing Factors" section below installed, may become extremely low
on memory causing other processes running on the system to be affected. It
may also be possible that "scdpmd" will dump core when it can't allocate
memory.

Workaround:
To work around this issue, disable "reboot_on_path_failure" property by using
the following command:
    # clnode set -p reboot_on_path_failure=disabled phys-foo-1

After changing this property, the load has reduces to almost 0, and the core
dumps should also stop.

Thanks!!!!


On Tuesday 26 February 2008, Markus Mayer wrote:
> Hi all,
>
> Hoping someone might be able to help here with two problems I have with
> scdpmd on a fresh Sun Cluster 3.2 installation on Solaris 10, update 3,
> with all patches installed that were available up until 15 February.  The
> systems have been "JASS'ed".
>
> First problem is that I have noticed that the scdpmd disk path monitor is
> eating on a V445 with 4 CPU's typically 22-25% of cpu resources.
> Essentially a whole cpu is being used just for disk path monitoring.  Right
> now there are 30 disks assigned to the system from a Hitachi AMS500, four
> paths to each disk, multipathing managed by MPxIO.
>
> The second problem is that scdpmd is dumping core quite often, at least
> every couple of days.  It does get restarted by the cluster infrastructure
> however the fact that is dumps bothers me.
>
> Does anyone know if there is any way to reduce the load, particularly as in
> the future I will get a lot more disks from storage assigned to the
> cluster, which would likely mean that the whole system resources would be
> eaten just with disk path monitoring.  According to the scdpmd manual, the
> monitoring could be shut off, however I have no idea what side effects that
> might have on the system or the cluster.
>
> Also, does anyone know anything that might help in resolving the core dumps
> that are happening with scdpmd.
>
> Thanks and regards
> Markus
> _______________________________________________
> sunmanagers mailing list
> sunmanagers@sunmanagers.org
> http://www.sunmanagers.org/mailman/listinfo/sunmanagers
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers