Re: ontape -p produce "Segmentation Fault" (IDS9.40UC5 on Solaris)

Hi, the info available in this case doesn't help a lot with diagnosis. But that is not your fault. Do you have a debugger installed on the machine ? If not, can you get one installed ? Standard "dbx" should be OK. With tha debugger installed you can a) check whether ontape has written a "core" file (look for a file named "core" in the directory where you executed "ontape -p" command that failed, and check the date of the file). If there is a core file, then you can use it for "post mortem analysis", like: "dbx $INFORMIXDIR/bin/ontape core" when you have the debugger's prompt you type a command like "where" or "backtrace". This will give you a so-called "function call stack trace". And this is what we need. Note: This post mortem analysis should always be done on the machine where the core file was written. Doing it on a different machine often produces a false stack trace. b) when there's no core file, you have to run the ontape command directly in the debugger. Start the debugger like this: "dbx $INFORMIXDIR/bin/ontape" when you get the debugger's prompt start the program run like this: "run -p" Then do everything as normal (answering ontape prompts, etc.). When the error happens, the debugger will report so. Then do a "where" or "backtrace" command to get the stack trace (as in a) above). If the problem does not reproduce (i.e. ontape works fine), repeat the whole procedure until you get a failure ... Keep the backup file with which you are able to reproduce the problem for possible later reproductions ! It is not sure that from the stack trace alone we can tell what the problem is, but it is the starting point. You may then have to open a case (aka "pmr") with IBM Informix Tech Support ... Regards, Martin -- Martin Fuerderer IBM Informix Development Munich, Germany Information Management owner-informix-list@iiug.org wrote on 17.03.2005 04:42:51: > > We installed IDS9.40UC5 and converted the database from IDS9.30UC5. The > conversion was successful. It only took about 50 seconds to finish the > conversion so it's obvious that not much update has taken place during > conversion. > > After the system is online with IDS9.40UC5 I took a level zero archive > into a disk file, which is about 80GB in size, then use "ontape -p" to > restore. With exactly the same archive, the physical restore sometime > works sometimes doesn't !! > > It always prints the chunk information, ask "Continue restore? (y/n)" > then "Do you want to backup the logs?", after a few seconds, > "Segmentation Fault" is displayed on the screen, ontape process has > gone but "oninit -r" processes are still running. When I use "ipcs" > command to monitor it, new shared memory segments are created after > "Segmentation Fault" has occurred. The "Physical restore ... started" > message has Not appeared in the engine log. No core dump is produced in > the directory where I ran "ontape -p". > > Sometimes it does work and everything is normal. Note again that on the > same archive sometimes it works sometimes it doesn't! > > I have got the latest Solaris recommented patch cluster installed by > the Unix admin but the problem still persist. > Also as suggested by IBM support I have tried run the "installserver" > and "RUN_AS_ROOT.server" scripts again, it didn't help. > > Any ideas? Your help is appreciated! > > > OS version is Solaris 5.8: > > misaccshm@osun9071:/mis/a00/users/informix/upgrade2004/archecker)uname > -a > SunOS osun9071 5.8 Generic_117350-21 sun4u sparc SUNW,Ultra-80 > > Screen cut and paste: > Continue restore? (y/n)y > Do you want to back up the logs? (y/n)n > Segmentation Fault > > > > Regards, > Carl Wu > sending to informix-list