Fwd: RE: semaphore tuning
Posted in 2005
--0-1358380275-1106270679=:40866
Content-Type: text/plain; charset=us-ascii
Floyd Wellershaus <fwellers@yahoo.com> wrote:Date: Thu, 20 Jan 2005 17:23:41 -0800 (PST)
From: Floyd Wellershaus
Subject: RE: semaphore tuning
To: "Sebastian, Norma J."
Thank you for your reply.
We had to go back to the 32 bit version because there was an incompatibility with some user defined function binaries ( not sure, it was a development issue ).
Below is the output of the af file before the semctl errors:
======================
14:17:56 IBM Informix Dynamic Server Version 9.40.UC4 Software Serial Number AA
A#B000000
14:17:56 Assert Failed: No Exception Handler
14:17:56 Who: Session(91, informix@hercules, 429, 3f2fdad0)
Thread(117, sqlexec, 3f2da558, 3)
File: mtex.c Line: 431
14:17:56 Results: Exception Caught. Type: MT_EX_OS, Context: mem
14:17:56 Action: Please notify IBM Informix Technical Support.
14:17:56 Stack for thread: 117 sqlexec
base: 0x40d1d000
len: 69632
pc: 0x007201c4
tos: 0x40d2cb98state: running
vp: 3
0x0071f5a8 (oninit)afhandler(0xb88400, 0x9fd000, 0x3f00a638, 0x1, 0x601, 0x601)
0x0071ed9c (oninit)afcrash_interface(0xafb4b4, 0x40d2d328, 0x3f00a638, 0xafb4cc, 0x1
af, 0x0)
0x0072370c (oninit)mt_ex_throw_sig(0x9fbfd0, 0xaf8644, 0xb859e0, 0xafb488, 0x3fc32e8
8, 0x0)
0x006f01f8 (oninit)afsig_handler(0xa, 0x40d2d900, 0x40d2d648, 0x0, 0x0, 0x6)
0x0029d888 (oninit)dup_ref_cmpseqs(0x40abe668, 0x404f6858, 0x0, 0x0, 0x4, 0x0)
0x002955dc (oninit)bldrefs (0xb9274c, 0x2952e4, 0x4000000, 0x404f6858, 0x3fd31be0, 0
x3fd31be0)
0x00356c8c (oninit)maybe_bldrefs(0x3fd31a40, 0x3fd31a00, 0x3fd31a40, 0x0, 0x3fd31a40
, 0x0)
0x00356aac (oninit)bldstructs(0xb92400, 0x20000, 0x1000, 0x10ca, 0x0, 0x40d2db64)
0x003567ac (oninit)sqcmd (0xb92400, 0xb9274c, 0x1, 0x4089c018, 0x3fd31a40, 0x0)
0x003563b8 (oninit)sq_cmnd (0xb92400, 0x0, 0x400, 0x6, 0x1000, 0xb76480)
0x003d38e4 (oninit)sqmain (0x1, 0xb92400, 0x9fac00, 0x1020, 0xb7f97c, 0x1000)
0x00741a80 (oninit)listen_verify(0xa2ed10, 0xa, 0x9faef0, 0x0, 0x40a76350, 0xafe6d0)
0x006fd960 (oninit)startup (0xb885a4, 0x7, 0x0, 0x3f00c8e0, 0x15, 0x3f00c8a0)
0x006f63fc (oninit)mt_poll_yield(0x0, 0x0, 0x0, 0x0, 0x0, 0x0)
0x00000000 (*nosymtab*)0x0
14:17:56 See Also: /informix/logs/tmp/af.45db264
---------------------------------Begin System Alarm Program Output
---------------------------------
Assertion Failure Type: CRASH
Host Name: vceb0a
Database Server Name: uolvceb0a
Time of failure: Wed Jan 19 14:17:56 EST 2005
AF file: /informix/logs/tmp/af.45db264
Shared memory file: None
System Blocking: OFF
===========------------- - - - - - -
onstat - output shows 1024000 Kbytes
and onstat -g seg shows
6000 1381386241 a000000 889192448 242472 R* 215096 1992
6001 1381386242 3f000000 159383552 5504 V 9580 29332
Total: - - 1048576000 - - 224676 31324
Looking at this, it looks like informix is only using 1 gig of the 8gigs available on the box. I guess I should bump up some buffers. We haven't allocated any extra segments yet, so I am assuming the initial size of 889 megs is sufficent.
Thank you.
"Sebastian, Norma J." <NormaJean.Sebastian@tellabs.com> wrote:
Hi Floyd,
>>Informix9.4-UC4 on Solaris 9
UC4?... that's a 32Bit version.
you did not post the entire error from your online log, what happens before your series of errno=22?
while you have the system shmmax set to:
set shmsys:shminfo_shmmax=4294967296
an IDS 32 bit version will top out at 3.75GB of memory or something close to that (at least it did on IDS 7.31 versions).
I am guessing your engine wants more memory, but is not designed to go higher.... so upgrade to a 64bit version (IDS 9.4 with a F where the U is in the full version number).
i.e...check your "onstat -" and see what your total memory is. Ours is:
IBM Informix Dynamic Server Version 9.40.FC3XD -- On-Line -- Up 02:24:34 -- 5570560 KbytesThe 5.57 GB number at the end means my engine takes 5.57GB of memory. the 9.40.F.... means I am a 64 bit version.
the number on the 'onstat -' line is not exact, if you do a 'onstat -g seg' you get more specific:
onstat -g seg
IBM Informix Dynamic Server Version 9.40.FC3XD -- On-Line -- Up 02:25:56 -- 5570560 KbytesSegment Summary:
id key addr size ovhd class blkused blkfree
4097 1383352321 10a000000 1929379840 477568 R* 467823 3217
4098 1383352322 17d000000 3758096384 115344 V 63473 854031
4099 1383352323 25d000000 2097152 720 M 317 195
4100 1383352324 25d200000 2097152 720 M 309 203
4101 1383352325 25d400000 2097152 720 M 309 203
4102 1383352326 25d600000 2097152 720 M 309 203
4103 1383352327 25d800000 2097152 720 M 309 203
4104 1383352328 25da00000 2097152 720 M 309 203
4105 1383352329 25dc00000 2097152 720 M 309 203
4106 1383352330 25de00000 2097152 720 M 309 203
Total: - - 5704253440 - - 533776 858864
(* segment locked in memory)
The size column here sums to 5.7GB whereas the "onstat -" showed 5.57GB. The classes in the above output are R=resident, V=Virtual, M=message.
Hope this info helps you with your situation,
Norma Jean
---------------------------------
From: owner-informix-list@iiug.org [mailto:owner-informix-list@iiug.org] On Behalf Of Floyd Wellershaus
Sent: Thursday, January 20, 2005 4:59 PM
To: informix-list@iiug.org
Subject: semaphore tuning
We are running Informix9.4-UC4 on Solaris 9.
We are getting intermittent semctl: errno = 22 errors that are causing assert failures.
Our current kernel settings for semaphores are:
set semsys:seminfo_semmap=64
set semsys:seminfo_semmni=4096
set semsys:seminfo_semmns=4096
set semsys:seminfo_semmnu=4096
set semsys:seminfo_semume=64
set semsys:seminfo_semmsl=500
set shmsys:shminfo_shmmax=4294967296
set shmsys:shminfo_shmmin=100
set shmsys:shminfo_shmmni=500
set shmsys:shminfo_shmseg=100
Is there any recommendation to determine what to up the semaphore values to ? Which value is it, is it the semmns value ?
Thanks,
Floyd
========================
-<<Floyd Wellershaus>>-
Database Administrator
Unix Administrator
email: fwellers@yahoo.com
Work: 703-733-4126
Pager: 703-705-9241
Email Pager: 7037059241@my2way.com
Home: 703-430-0805
Cell: 703-477-6045
========================
============================================================The information contained in this message may be privilegedand confidential and protected from disclosure. If the readerof this message is not the inten