[not newbie] Internal heap ERROR and core with 8i on Linux

From: Michael Ju. Tokarev <mjt_at_tls.msk.ru>
Date: Thu, 07 Oct 1999 21:25:32 +0400
Message-ID: <37FCD78C.DCF8E342@tls.msk.ru>

Hi, All!

This is "my last step" -- to make a question here. I installed 8i (EE, taked from ftp.oracle.com) on RedHat linux. Yes, there are many gotchas during install, I was able to resolve them all. Now it is up and running well (at the first look at least). But before I encountered some problems with Oracle Forms client. Some queries from forms6.0 craches the server process and it generates trace and core files. Error is 'Internal heap ERROR 17182', as stated in .trc file, and here is some excerpt from it:

HEAP DUMP heap name="session heap" desc=0x2217d330 extent sz=0x3eec alt=32767 het=32767 rec=0 flg=2 opc=3 parent=20000134 owner=2005be88 nex=0 xsz=0x3eec EXTENT 0

  Chunk 245121d8 sz=     2692    free      "               "
  Chunk 24512c5c sz=       76    freeable  "frame segment  "
  Chunk 24512ca8 sz=     2008    freeable  "frame segment  "

[...]
Chunk 2451c628 sz= 16867072 ERROR, BAD MAGIC NUMBER (1015f00) Total heap size = 16909080
[...]
ksedmp: internal or fatal error
ORA-00600: internal error code, arguments: [17182], [609338920], [], [], [], [], [], [] Current SQL statement for this session: select ROLE GRPNAME ,FLAG GRPFLG from FRM50_ENABLED_ROLES order by ROLE desc [...]
Dump of memory from 0x0 to 0x40 (!)
...
ub4 ksmugmg2 [2217DE64, 2217DE68) = 0000CDCD Exception signal: 11, code = 927cf20
ORA-07445: exception encountered: core dump [11] [153603872] [] [] [] [] ....

This is an internal error, I know, and it should be reported as a bug -- but I'm not a Oracle customer for 8i -- this is just a testdrive for me (we are using 7.3 on Solaris now).

But problem is more "interesting":

server always crashes on this SQL stmt;
there is one (!) client for that server always worked fine (!), all other client machines are "bad";
if I make some queries in forms, problem does not appears until end of forms session
other queries caused forms to crash server (but if I make before them others as in prev. note, all is ok)
there is not matter that oracle/linux/hardware configuration is used -- I tried this with fresh redhat 6.0, with all patches, with 6.1, with original oracle executables from .jar files, with relinked ones, without even java "option", with many different config parameters and shmem parameters, and on three different machines -- all the same. I also tried that with "unpatched" oracle, with original patches and also with SuSe patches (excellent work!).
shared/dedicated server config also does not matter.
hardware config also does not mattes - it works on 32meg and on 256meg and crashes always the same.

Tthis seemed to mee that this is really Oracle bug, not my mistake. And this seemed to me that there are some uninitialized variable in Oracle server that caused this trouble. The only one statement that seemed to crash server (it is ok from sqlplus) is about roles. We using many roles for our big task, and I set max_enabled_roles in init.ora to 100 (testdrive only!). I can't reproduce this error if I leave parameter with default value (but my forms does not work -- I need to change code so it doesn't depend on roles). Maybe this is a point? I never see troubles with that earlier (on 7.3 we have max_enabled_roles = 300, and really have about 250 enabled ones).

And also I can't make another client machine that also will be "good" for server. I really can't understand what is the difference between that "good" client and the others (all of them are win95 machines).

Does anyone have a solution/workaround? Does this a known bug? Maybe such situation exists on other platforms?

We have a customer that already wants to buy Oracle on Linux, but if this problem exists we should not recommend to do so, at least now...

Thank you!

Regadrs,
  Michael Ju. Tokarev,
  System and Database Administrator,
  JSC "Telecom-Service",
  Moscow, Russia. Received on Thu Oct 07 1999 - 12:25:32 CDT