PROTCONV crashes every Monday morning

[Former Member]

Hello,

I've the problem on my SLES10 with MaxDB 7.7.07.24 that the tool protconv every monday at ca. 04:30 crashes with segfault. The only reason I know this, is our monitoring system wich alerts the following text from /var/log/messages:

Feb 14 04:26:34 hk-bhbp0 kernel: protconv[32369]: segfault at 0000000000000000 rip 00002ada01206990 rsp 00007fffaa2ed2b8 error 4

The Problem occured first time, after I had updated the DB from 7.7.06.15 to 7.7.07.24. Does anybody has an Idea what's happening here? If I start protconv manually the entry in /var/log/messages does not appear.

Does protconv only runs one time per week?

Thanks in advance for your answers.

Norman

[thorsten_zielke · Accepted solution]

Hi,

the purpose of 'protconv' is to convert the 'pseudo-XML' error log files in /sapdb/data/wrk/<instance> to plain text [(for details click here)|http://wiki.sdn.sap.com/wiki/x/PwErAg].

Basically, protconv is used by other applications like Database Studio, Database Manager or R/3 transaction DB50 or via dbmcli whenever the log files are requested.

I have not heard of any segmentation fault in combination with protconv, but to proceed here, I would suggest determining the application that calls protconv at 04:30. If this is an R/3 system, please check for any database jobs around that time. Also, is there any corresponding error message in the 'dbm.prt' file?

Regards,

Thorsten

[Former Member · Accepted solution]

Hi,

sorry for the delay and thanks for this answer. I had a look through the Database Manager Log File and found some corresponding lines:

2011-04-11 05:02:10 0x0000253d INF 1 DBMSrvCo A DBM Server client connection was established at 2011-04-11 05:02:07 (client process has process ID 9525 on computer hk-bhbp0.hk-sap.os.itelligence.de).

2011-04-11 05:02:10 0x0000253d INF 421 DBMSrv Command 'file_getfirst ...' was executed since 2011-04-11 05:02:07.

2011-04-11 05:02:10 0x0000253d ERR -24580 DBMSrv ERR_COMMAND_FAILED: Command 'file_getfirst' has ended and failed with return code -24964.

0x0000253d ERR -24964 DBMSrv ERR_EXECUTE: error in program execution

0x0000253d ERR -24778 DBMSrv 35584,/sapdb/programs/bin/protconv -p /sapdb/data/wrk/HBP -f classic -t _KNLMSGARCHIVE -o /sapdb/data/wrk/HBP/.KnlMsgArc.txt

2011-04-11 05:02:10 0x0000253d INF 2 DBMSrvCo A DBM Server client connection was released (client process has process ID 9525 on computer hk-bhbp0.hk-sap.os.itelligence.de).

Until now, I could not trace back which client process have had the process ID 9525. Maybe you can - with the log information above - imaging what is connecting to the DB and what is executing the command 'file_getfirst'

Regards,

Norman

[thorsten_zielke · Accepted solution]

Hi,

sorry for beeing late myself.

Looking at the dbm.prt logfile, the following statement has triggered the 'protconv'-crash:

'/sapdb/programs/bin/protconv -p /sapdb/data/wrk/HBP -f classic -t _KNLMSGARCHIVE -o /sapdb/data/wrk/HBP/.KnlMsgArc.txt'.

Please execute just this line and let me know the result.

[Former Member · Accepted solution]

Hello,

I tried to execute the statement manually and got the same error:

Segmentation fault

Also in the message file the error was logged: kernel: protconv[19012]: segfault at 0000000000000000 rip 00002b11e41ad990 rsp 00007fffc73459f8 error 4

Anyway, the file was created correct.

So now, what can I do to prevent this? Is the protconv tool buggy?

Kind Regards,

Norman Stoer

Edited by: Norman Stoer on May 2, 2011 11:10 AM

[thorsten_zielke · Accepted solution]

Hmm, as protconv is called with the '-t _KNLMSGARCHIVE' option, it would try to translate the 'KnlMsgArchive' file from pseudo-xml to test. Maybe there is a problem with your current 'KnlMsgArchive' file?

Please test that and let me know the result. What you can do is to

1. If you can afford a downtime, stop the database, rename the 'KnlMsgArchive' file, restart and retry to crash protconv.

2. If you have another MaxDB database at hand, try the same 'protconv' command to that other DB (but change the database name and paths to match that instance).

Thorsten

[Former Member · Accepted solution]

Hi,

I tried to execute on another system but did not get the segfault. Unfortunately I had no downtime for the system until now so I could'nt change the knlmsgarchive-file.

But I tested some things and found out, that probably the switch -f (output format) in the command is the problem. I checked the two other options (plain and xml) for the switch an got no segfault.

How the filekey (switch -t) was written, means with underscore or not, was not relevant.

Next time I get a downtime I will try to make a new knlmsgarchive-file and check wether the protconv command will work or not.

Regards,

Norman Stoer