when informix-server down ?
Hi,- I think it`s better when I bring the informix db-server from time to time down and start the server new. My question: Can I see when it`s the right time (memory, ..)
In article <38560b8b.2542776@news.btx.dtag.de>,
him_lommel@t-online.de (Andreas Lommel) wrote:
> Hi,- I think it`s better when I bring the informix db-server from time
> to time down and start the server new. My question: Can I see when
> it`s the right time (memory, ..)
>
>
The only time that I noticed a performance degredation caused by time
that was fixed by bouncing the instance was when I was working on a
server that used process aging. The OS was lowering the priority on the
database because they had been up for a long time. To fix this, you can
set NOAGE to 1 in the config file, I believe. Look it up in the book to
make sure, I don't have one handy. Other than that, it's possible that
over time you are allocating extra shared memory segments and bouncing
the instance drops it back to one. You can see this by doing an ipcs -a
or onstat -g seg, I believe. If this is the case, you might want to
increase SHMVIRTSIZE to whatever you are using when you see degredation.
--
# unrm /
ksh: unrm: not found
# man cpio
Sent via Deja.com http://www.deja.com/
Before you buy.
Andreas Lommel wrote:
>
> Hi,- I think it`s better when I bring the informix db-server from time
> to time down and start the server new. My question: Can I see when
> it`s the right time (memory, ..)
PLATFORM and VERSION info?
There should be no need to bounce the engine periodically. If you are
running on Solaris or HP you should set NOAGE to 1 to prevent the OS from
reducing the priority of the oninit processes over time. If you are on HPUX
you need to make sure there are no more than two (and preferrably only one)
virtual memory segment as HP processors bog down if you try to access more
than four shared memory segments and the Resident, initial Virtual, and
shared memory communications segments are three so any additional virtual
segments puts you up against the wall and performance drops to the floor
if another virtual segment is allocated or if your application uses shared
memory. On HP fold the total of all virtual segment sizes into SHMVIRTSIZE,
increase SHMADD, and bounce the server to reduce the number of virtual
segments.
On other platforms too much shared memory allocations can slow things a bit
so run onmode -F periodically.
Art S. Kagel