Online 7.12 and 7.20 on SCO OpenServer 5 - CRASH - HELP

This is a multi-part message in MIME format. ------=_NextPart_000_01BB9F69.29F19020 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Wel, Informix let me down again. About a week ago I posted a message about crashes of Online 7.12 UC1 and 7.12 UC2 running under SCO Openserver 5 This was the message : I have three Compaq Machines running SCO Openserver 5 release 5.00b Compaq ProSignia, 64 Mb, Informix 7.12 UC1 development Compaq ProLiant 4500, dual 133 Mhz Pentium, 64 Mb, 7.12 UC2 run-time Compaq ProLiant 5000, dual 166Mhz Pentium Pro, 128 Mb, 7.12 UC2 run-time In four days time al three engines crashed with the same error in the online.log file. A problem with shared memory followed by bad magic numbers, problems with a memory allocation in a certain C - source and then boom. Engine gone. DB fucked up. Nice. This happend on al 3 servers. It happend on the ProLiant while performing an alter table using dbaccess and on the ProSignia while performing a select. All tables accessed during the crash are FUBAR. Oncheck reports constraint problems and index-corruption. Informix Belgium informed me that this could be a bug in release 7.12, sometimes experienced on SCO and HP-UX platforms. Now the follow up : Informix Belgium advised me to upgrade to Online 7.13. This release doesn't exist in a SCO Openserver 5 version so I had to upgrade to Online 7.20 UC1. A lot of people responded to the above (original) message advising me to upgrade SCO to version 5.00d. I discovered that the versions of SCO I was running where indeed SCO OpenServer 5 release 5.00b whith the Release supplement 5.00d installed. In plain English : I was running SCO Openserver 5.00d (I love SCO system admin tools, they are really good ...). I used the ProLiants as test machines for the upgrade, running the following config : Compaq ProLiant 4500, dual 133 Mhz Pentium, 64 Mb, 7.20 UC1 run-time Compaq ProLiant 5000, dual 166Mhz Pentium Pro, 128 Mb, 7.20 UC1 run-time Both machine are running SCO OpenServer 5 release 5.00b with the followin patches installed Maintenance Supplement rs500d (upgrade to release 5.00d) Networking Supplemt 1.0 (fixes bugs in TCP/IP) HTFS supplement oss434a (fixes bugs in the HTFS file-system) Kernel Error Path supplement oss436a (fixes security holes) Streams Lock Timeout & Protocol Data supplement oss437a (fixes bugs in streams implementation, also required by Informix High-Performance Loader) SUDS aio supplement oss441a (fixes bugs in SUDS driver) The aio-driver and suds-driver are linked into the kernel. I installed Online 7.20 UC1 on both machines. I configured a db on the ProLiant 5000, tried an alter table and boom. Same error, same crash. Online 7.20UC1 does a better job at handling the crash. It produced a kind of dump. Both the dump and the online.log are included with this message. I also noticed a strange behaviour concerning KAIO. The release notes mentioned the KAIO is supported in release 7.20 (new feature) but it has to be enabled by setting an environment variable. This does NOT work on the ProLiant 5000 (no messages). It seems to do something on the ProLiant 4500 (some UNIX error messages concerning availability of shared memory - SCO locks KAIO shared memory) but KAI is NOT enabled. Any ideas ? -- Koen ------=_NextPart_000_01BB9F69.29F19020 Content-Type: application/octet-stream; name="af.438b9d" Content-Transfer-Encoding: quoted-printable Content-Description: af.438b9d (438B9D File) Content-Disposition: attachment; filename="af.438b9d" 10:16:39 Found during mt_shm_free 1 10:16:39 Pool '50' (0x25b2014) 10:16:39 Bad block header 0x26baf28 blk-64 026baee8: d0b06b02 00000000 fcae6b02 63646766 ..k..... ..k.cdgf=20 026baef8: 00000000 54b16b02 00000000 10af6b02 ....T.k. ......k.=20 026baf08: 63646b64 00000000 d8b16b02 00000000 cdkd.... ..k.....=20 026baf18: 5cb26b02 63646b62 00000000 a89e6b02 \\.k.cdkb ......k.=20 blk+64 026baf28: a89e6b02 a89e6b02 a89e6b02 a89e6b02 ..k...k. ..k...k.=20026baf38 * 026baf48: a89e6b02 a89e6b02 a89e6b02 5f616c67 ..k...k. ..k._alg=20 026baf58: 5f636465 5f697076 5f666b00 b8ae6b02 _cde_ipv _fk...k.=20 10:16:39 Bad block removed from pool 10:16:39 =20 10:16:39 Assert Failed: Memory block header corruption detected in = mt_shm_free 1 10:16:39 Who: Session(50, informix@civcti, 1413, 0) Thread(67, sqlexec, 0, 1) 10:16:39 Results: Pool repaired 10:16:39 Action: Please notify Informix Technical Support. 10:16:39 See Also: /tmp/af.438b9d 10:16:39 Stack for thread: 67 sqlexec base: 0x026939b8 len: 36864 pc: 0x00260a31 tos: 0x0269c120 0x0269c184 ***unknown*** (0x2693864, 0x0, 0x43db20, 0x25b2014) 0x0026047e mt_affail (0x269c56c, 0x43db20, 0x25d9d00, 0x0) 0x002597ef recover_pool_bad_block(0x25b2014, 0x26baf28, 0x43d27c, = 0x26b9ea8) 0x00257e8f mt_shm_free (0x25b2014, 0x26baf3c, 0x25e15a8, 0x5) 0x002264f2 do_rhpfree (0x26b801c, 0x25b2014, 0x25e1494, 0x269c878) 0x002264c0 rhpfree (0x26b801c, 0x25e1494, 0x26b0560, 0x1) 0x000e4b89 releasesdb (0x25e1494, 0x1, 0x26b043c, 0x1f) 0x00024cc9 sqdbclose (0x1, 0x25b2170, 0x269c8ac, 0x1f) 0x000109bc aud_sqdbclose (0x400948, 0x0, 0x26b043c, 0x26b043c) 0x000d7255 excommand (0x26b043c, 0x400948, 0x4, 0x7) 0x000bcb74 sq_execute (0x0, 0x464328, 0xa, 0x25d7950) 0x000ca17f sqmain (0x0, 0x0, 0x0, 0x0) 0x0024b3ec startup (0x25, 0x269ca04, 0x26939a4, 0x4c) 0x00000000 ***unknown*** =20 10:16:39 =20 ------------------ End of assertion failure 0 ----------------- 10:16:39 Found during mt_shm_free 1 10:16:39 Pool '50' (0x25b2014) 10:16:39 Bad block header 0x26b9e94 blk-64 026b9e54: 44a36b02 7800f008 409e6b02 00000000 D.k.x... @.k.....=20 026b9e64: 00000000 80a36b02 02000000 4c886b02 ......k. ....L.k.=20 026b9e74: 4c816b02 00000000 00000000 74ad6b02 L.k..... ....t.k.=20 026b9e84: 00000000 00000000 15000000 28a36b02 ........ ....(.k.=20 blk+64 026b9e94: 8c9a6b02 9c040000 5042d704 a0966b02 ..k..... PB....k.=20 026b9ea4: 00000000 6465745f 616c675f 75690000 ....det_ alg_ui..=20 026b9eb4: 00000000 00000073 69646973 00000000 .......s idis....=20 026b9ec4: 01000200 03000400 00000000 00000000 ........ ........=20 10:16:39 Can't find corrupted block's previous block 10:16:39 =20 10:16:39 Assert Failed: Memory block header corruption detected in = mt_shm_free 1 10:16:39 Who: Session(50, informix@civcti, 1413, 0) Thread(67, sqlexec, 0, 1) 10:16:39 Results: Unable to repair pool 10:16:39 Action: Please notify Informix Technical Support. 10:16:39 See Also: /tmp/af.438b9d 10:16:39 Stack for thread: 67 sqlexec base: 0x026939b8 len: 36864 pc: 0x00260a31 tos: 0x0269c120 0x0269c184 ***unknown*** (0x2693864, 0xffffffff, 0x43db4c, 0x25b2014) 0x0026047e mt_affail (0x269c56c, 0x43db4c, 0x25d9d00, 0x0) 0x002597ef recover_pool_bad_block(0x25b2014, 0x26b9e94, 0x43d27c, = 0x5f746564) 0x00257e8f mt_shm_free (0x25b2014, 0x26b9ea8, 0x25e15a8, 0x5) 0x002264f2 do_rhpfree (0x26b801c, 0x25b2014, 0x25e1494, 0x269c878) 0