[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] RHEL3 Cluster Broken Pipe error and Heartbeat configuration



Hi Lon,

   Thanks a lot for your valuable reply, I am waiting for your reply
only, i know this type of cluster error i got only during heavy I/O to
the shared storage(scsi) even i read the sar report telling that
system  idle percentage of merry 1 to 2 % at that time.If i go with
fibre channel storage can i prevent this type of issues without
upgrading the clumanager version. Also Please help me in configuring
the below part if your suggesting some thing else for the below thing
than also i will be very happy.

############################################################
 Also  anyone help me to configure a dedicated LAN (for example eth3)
as heartbeat(private  point to point cross over cable network for
cluster communications),I don't wish heartbeat over public LAN ,
because of heavy Network saturation.

 Forthe above heartbeat configuration  i didn't found any suitable
document for RHEL. Can  you provide me the suitable link or guide
me what are all the changes i have to made in my  existing cluster.xml
 file for this private heartbeat configuration to work.
#############################################################

Anticipating your reply


On Sat, Nov 15, 2008 at 3:24 AM, Lon Hohberger <lhh redhat com> wrote:
> On Wed, 2008-11-12 at 19:14 +0530, lingu wrote:
>> cluquorumd[1921]: <warning> Disk-TB: Detected I/O Hang!
>
> Eep.
>
> This means that I/O to shared storage has gotten slow.  Strange.  I
> heard reports of this on another cluster (after going from U3->U8), but
> I don't know what the cause is.  With this cluster, we straced the
> cluquorumd process and found that it was slowing down *a lot* in the
> write() call when writing to shared storage.
>
> You can try the current U9+erratum clumanager or the test release if you
> want to (it makes unlock more robust when I/O performance is slow for
> some reason).
>
> However, someone really needs to profile the kernel if you're seeing
> slow write times while stracing cluquorumd...
>
> -- Lon
>
> --
> Linux-cluster mailing list
> Linux-cluster redhat com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]