RE: Oracle on AIX hangs

Brandon,

Hdisk0 on AIX (based on my memory from 5 - 6 years ago) normally houses the OS including the swap area. Additionally, the file I/O buffer is normally memory-unbounded and can take up precious memory. If you are using a larger memory footprint and inefficiencies/incorrect SQL is forcing a lot of I/O, then you might actually be suffering from excessive paging and possibly even swapping. Check sar -rq will reveal this (r = memory, q = CPU/Swap queue).

Hth,
John Kanagaraj <><
DB Soft Inc
Phone: 408-970-7002 (W)  

Co-Author: Oracle Database 10g Insider Solutions http://www.amazon.com/exec/obidos/tg/detail/-/0672327910/  

• The opinions and facts contained in this message are entirely mine and do not reflect those of my employer or customers **

-----Original Message-----
From: oracle-l-bounce_at_freelists.org [mailto:oracle-l-bounce_at_freelists.org] On Behalf Of Stephane Faroult Sent: Thursday, January 26, 2006 1:25 PM To: Brandon.Allen_at_OneNeck.com
Cc: oracle-l_at_freelists.org
Subject: Re: Oracle on AIX hangs

Brandon,

   Knowing what you have on hdisk0 might be useful. Would you by chance use shared servers? I have met a remotely similar case with shared servers and a shared pool too small (even if the large pool was suitably set), but it was 8.1.7 on Solaris and the problem was fixed in 9.  I have trouble linking excessive disk usage to slow telnet connection.

Stéphane Faroult

On Thu, 2006-01-26 at 13:22 -0700, Allen, Brandon wrote:
> Anyone else seen anything like this?
>
> Oracle 9.2.0.6, AIX 5.2 ML7
>
> Recently migrated to new server, from Oracle 8.1.7 & AIX 5.1, and
> since moving to the new server we've had a few instances of horrible
> performance where hdisk0 is pegged at 100% utilization. When this
> happens it takes 60 seconds just to establish a telnet connection and
> all Oracle sessions, as well as new connection attempts just hang
> indefinitely. We can't even get a "conn / as sysdba" connection so we
> can't get a hanganalyze or systemstate dump.
>
> After we killed all database server processes (not the background
> processes), the hdisk0 utilization dropped down to zero, but then as
> soon as we try to connect again, the connection attempt hangs and
> hdisk0 goes back to 100%. We can't even shutdown the instance since
> we can't connect so we have to kill it, and then the problem goes away
> and does not come back when the instance is restarted . . . until the
> next time, which may be days or weeks later. We haven't been able to
> identify any pattern to the occurrences yet.
>
> I've been all over Google and Metalink with no luck.
>
> Any ideas?
>
> Thanks,
> Brandon

--
http://www.freelists.org/webpage/oracle-l


--
http://www.freelists.org/webpage/oracle-l