Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Mailing Lists -> Oracle-L -> Re: RAC 10.2.0.3

Re: RAC 10.2.0.3

From: goran bogdanovic <goran00_at_gmail.com>
Date: Sun, 18 Nov 2007 10:13:32 +0100
Message-ID: <6d0a3ba80711180113q69b08a6anb7f178acfcf48519@mail.gmail.com>


I can't say how MTU size is managed on HP-UX...but at least AFAIK in Linux should not be an issue, because Linux implements a method known as "path MTU discovery" (if enabled) to determine the MTU on a site-by-site basis. If the initial (or manually adjusted) MTU used for a connection is too high, Linux throttles it back until it works...kind of self optimizing the connection.
Maybe you should speak with your SA to check what is behavior in HP or someone on the list with better HP knowledge can give you more info.

regards,
goran

On Nov 15, 2007 4:11 PM, Rajeev Prabhakar <rprabha01_at_gmail.com> wrote:

> Hello Ujang
>
> You should look into voting disk tmeout settings for the relevant nodes
> and
> work with SAs/Storage Admins to arrive / ascertain the new values.
> Ideally,
> you should stress/load the database to determine the iops levels at which
> is experiences a meltdown.
>
> Also, verify the mtu settings via tracepath and determine the packet loss
> /
> and/or transmission rate if you increase the mtu settings. Jumbo frames
> might be an option assuming all the intermediate hop points support those
> settings.
>
> Review the hot blocks and look for concurrent jobs that could be
> contributing
> to such scenarios. If you can, stagger them, it would help. Look into any
> options to partition the relevant objects
>
> -Rajeev
>
> On Nov 15, 2007 12:11 AM, Ujang Jaenudin <ujang.jaenudin_at_gmail.com> wrote:
> > all,
> >
> > i have 2 nodes RAC 10.2.0.3 on UX
> > in 8k blocksize, what is the best MTU size to be set? should I set 8k
> too?
> >
> > I have the problem with high disk utilization, which causing node
> eviction,
> > the FC is multipathing, so how is the good setting on misscount??
> >
> > about this parameter, how can help me???
> > because user experiencing some locks on gcs, cache buffer chains, etc.
> >
> > _gc_affinity_time=0
> > _gc_undo_affinity=false
> > _gc_affinity_limit=100000
> > _gc_affinity_minimum=1000000
> >
> >
> > --
> > regards
> > ujang
> > jakarta - indonesia
> > http://ujang-id.blogspot.com
> --
> http://www.freelists.org/webpage/oracle-l
>
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Sun Nov 18 2007 - 03:13:32 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US