Help! MaxDB 7.5 crash--how to diagnose the cause/resolution?

[marvina_boehm]

We have a Kpro server running MaxDB 7.5 (on Windows) as a Content Server for Engineering drawing vaulting. It crashed today and won't recover.

Errors in file "knldiag.err"

-

---

<BEGIN>

2007-11-12 11:56:13 --- Starting GMT 2007-11-12 17:56:13 7.5.0 Build 029-121-099-958

2007-11-12 12:01:43 0xD4C ERR 53118 CONFIG SegmentSize < 10 :; 5

2007-11-12 12:01:43 0xD4C ERR 18196 DBCRASH vabort:Emergency Shutdown, vgg01.c: 5392

2007-11-12 12:01:43 0xD4C ERR 19999 BTRACE SymbolSearchPath: c:\sapdb\data\wrk\SDB;C:\SAPDB\SDB\DB\pgm;C:\SAPDB\SDB\DB\symbols;C:\SAPDB\SDB\DB\sap;C:\WINNT;C

2007-11-12 12:01:43 0xD4C ERR 19999 BTRACE :\SAPDB\SDB\DB\sap\

2007-11-12 12:01:43 0xD4C ERR 19999 BTRACE -

---

> Register Dump <----

<END>

The most critical seems to be message 53119, but this doesn't seem to be documented anywhere!

Thanks for any help,

Marvin

[markus_doehr2 · Accepted solution]

I suggest you open an OSS call so that the support people can investigate, what went wrong.

What you post is just the final error, check knldiag where "ERR" appears the first time.

--

Markus

[former_member229109]

Hello Marvin,

You already created the OSS VH message 945486/2007 for SAP, where you reported this problem. Correct?

The knldiag was attached to the message & As you could check in the knldiag file::

...

2007-11-12 12:01:43 0xD4C 13 Pager Stop Read Converter, Pages: 13809 IO: 13809

2007-11-12 12:01:43 0xD4C ERR 53118 CONFIG SegmentSize < 10 :; 5

2007-11-12 12:01:43 0xD4C ERR 18196 DBCRASH vabort:Emergency Shutdown, vgg01.c: 5392

…

Before the database was Shutdown with 'vabort:Emergency ' it was stopped to read Converter, Pages:

13809 IO: 13809

As you could see in the knldiag file the current value of the CACHE_SIZE parameter is set CACHE_SIZE=14165 < pages, 8KB pro page >.

The MaxDB database SDB failed to start, because it could not allocate the memory it needed.

The CACHE_SIZE was set to 14165 pages (each page has a 8 KB size).

Now, the reason for this situation is that the data cache, the converter cache and other MaxDB memory structures are allocating memory from the same cache. The problem was that the converter cache is allocated during database start and the size depends on the database filling level, because the fuller the DB, the more pages need to be addressed from the converter cache.

As your cache size was configured quite small with only 14165 pages and probably the filling level has increased over the time, the converter cache allocated almost all of the whole database cache and therefore insufficient space was left for the other database structures, e.g. the data cache. However, MaxDB is only allowed to use the CACHE_SIZE memory for this cache.

Did you increase the CACHE_SIZE from 14165 pages to what value?

< From the knldiag file please check:

Max virtual memory: 2047 MB

….

Please see if you could use the 3GB option? >

-> Please update the OSS ticket 945486/2007 if you still have the problems to restart the database.

Please attach the new knldiag file to the ticket 945486/2007.

Thank you and best regards, Natalia Khlopina

[marvina_boehm]

We do in fact have a message open on the Support Portal. They will likely not response quickly, and we'll probably start a recovery from the last backup.

Good point about the last message in the log--the messages above are when we try to restart again (the server was restarted to try and clear things).

I looked all the way back in th logs for the very first error, and this is what shows:

2007-11-12 10:04:30 0x934 ERR 52000 CHECK KB05 fatal error; 1920

2007-11-12 10:04:30 0x934 ERR 18196 DBCRASH vabort:Emergency Shutdown, vgg01.c: 5392

Again, though message 52000 is not documented!

Marvin

[markus_doehr2]

KB points to "Kernel: Lock and log", KB05 seems to be

kb/gkb05: Locklist_constants_and_types

and

kb/vkb05: KB_functions

(if those interface descriptions from http://home.snafu.de/~dittmar/sapdbdev/sapdbModules.txt are still valid)

so it's a kernel internal function, that fails, thus you don't find that error message.

--

Markus

Message was edited by:

Markus Döhr