David, do you know what calculations, if any, to perform to arrive
with the number 50?
Our server has 256MB RAM. Page space is 512MB. The ADSM
server is the only application that is running in our SP wide node.
What would be the optimum number to use for vmtune given the
above?
Your suggestion is appreciated.
Angel
-----Original Message-----
From: bohm [SMTP:bohm AT US.IBM DOT COM]
Sent: Tuesday, August 31, 1999 1:56 PM
To: ADSM-L
Cc: bohm
Subject: Re: Server crashed - BUF043.
I do not think this is necessary. The BUF043 problem should be fixed
by
IX87338 which is included in the 3.1.2.30 level. The fix for the
BUF006
should be included in a future PTF through APAR IY00869.
An underlying problem to the BUF006 appears to be an I/O bottleneck
preventing
dirty pages from getting flushed fast enough. A circumvention to
that
problem that has worked for at least one customer is
vmtune -P 50
I would try this first. APAR IY00869 will fix the BUF006 problem,
but you
will still have the problem of memory utilization and the I/O
bottleneck
after that fix which the vmtune setting should resolve. If the I/O
bottleneck is resolved then you should never get into the condition
that causes the BUF006 problem which is a timing problem that
only occurs when the data base buffer pool is full of used pages.
David Bohm
ADSM server development
(520) 799-5082 T/L 321-5082
email - bohm AT us.ibm DOT com
>Short of an upgrade I'd suggest the following...
>
>DUMP/LOAD or BACKUP/RESTORE DB. PReferrably on top of a fresh
install
>of ADSM. Those BUF errors are uncommon on most systems, so yours is
a
>special case.
>They could probably run traces and perhaps rig a fix, but a fresh
install of
>ADSM
>may be the way to go if you want results anytime soon.
>
>PMR's can be reopened and crossref'd if need be.
>
>ADSML
--openmail-part-0dd16144-00000001--
=========================================================================
|
