sh....
ok, all is lost ...
I have a backup of /var/opt/sapdb so i copy the files back...
What is the command to reset the flag?
CU
Andreas
Mensing, Joerg schrieb:
Hi Andreas,
sorry, but with "init config" you erased the complete content of your database including the message tables.
The -118 means
SQLM.en: ENG -118 SERVERDB must be activated
From knldiag.err i got
2002-10-30 15:45:53 1542 ERR 11000 devio write error: Cannot allocate memory
----->>>>>> that is exactly the 'LINUX only' return code for the read/write problem due to LINUX kernel resource problem...
Your hard disk did not reported an error, so you do not have to change it.
Again with "init config" you erased the complete content of your database including the message tables. This leads to 'Message not available'. After an init config you restart from scratch. Now you should ever restore from backup or you have to activate
the empty database with
util_connect YOUR_USER,YOUR_PASSWORD
util_execute init config
util_execute activate serverdb sysdba YOUR_NAME_FOR_SYSDBA password YOUR_PASSWORD_FOR_SYSDBA
util_release
LOAD_SYSTAB -u YOUR_NAME_FOR_SYSDBA,YOUR_PASSWORD_FOR_SYSDBA -ud DOMAIN
db_warm
CU
jrg
-----Original Message-----
From: Andreas Ernst [mailto:ae@;ae-online.de]
Sent: Donnerstag, 31. Oktober 2002 09:31
To: Mensing, Joerg; SAPDB
Subject: Re: -24988 sql error [db_warm -f]; -902,Message not available
Hi J�rg,
thanks for you answer.
Mensing, Joerg schrieb:
The Message
2002-10-30 17:08:29 1671 ERR 11000 page0io_ devspace
marked as bad at pos 55722
shows that not just the vserver was killed, but an I/O error
occured on your data devspace before.
Look in your knldiag.err (in the same directory as knldiag).
I think you will find there the reason why
the devspace is marked as bad. If it was a physical I/O
error (EIO) you have to replace your harddisk (sounds
terrific doesn't it)
and do restore from your last backup. But since it is a
recent LINUX there may be another reason:
There are recently some Linux kernels which have an
undocumented feature... They report an I/O error for read or
write, not due to I/O error, but because the LINUX kernel
runs out of memory!!!! Instead of stalling the
This problem i have on this computer, i have a Java-App
runing on this
machine and this program is also killed!!
process until the internal resource is available again, the
error is returned to the user (in our case SAPDB). This
'feature' is undocumented (not in man read) and automatically
leads to 'marked as bad' since the SAPDB cannot distinguish
between 'normal' read/write error and internel LINUX kernel
problems (yet...). This running out of memory only happens in
a high load situation and the 'feature' is LINUX kernel
version dependend. Some kernel implementators put
responsibility for handling an out of memory situation to the
application. SAP Linuxlab gave me the information, that this
'behaviour' is a reason why SAP only certifies well tested
LINUX kernel version for usage with R/3.
Dont misunderstand me... If you have a program that can
handle a resource problem in the underlying kernel, this read
'behaviour' is a good idea, but 'LINUX only' changes in
behaviour of common system calls cannot be handled automatically....
So if you did not have 'just' an 'ordinary' I/O problem (but
ONLY if you are sure...), you can reset the 'marked as bad'
with a diagnose command in ADMIN (cold) mode and restart. If
you reset the flag, but you had an I/O problem, your database
will crash again with very high propability.
CU
jrg
Meanwhile t tried the command 'util_execute init config'...
Now i have the error '-24988 sql error [db_warm -f]; -118,Message not
available'
Do i have now less or more problems?
What is the command to reset the flag? Or is there now no
more change..
cu
Andreas
_______________________________________________
sapdb.general mailing list
[EMAIL PROTECTED]
http://listserv.sap.com/mailman/listinfo/sapdb.general
