USING Q4 TO ANALYZE SYSTEM DUMP FILES (For 10.10/20/30) --------------------------------------------------------- A crashing 10.X HP-UX system saves a snapshot of RAM in swap, reboots, and copies the resulting "dump" into /var/adm/crash. Follow this article to speedily analyze the dump and email the resulting ASCII text file to the HP Response Center. ============================ STEP 1 ============================ If /usr/contrib/bin/q4 exists, skip to step 2. If not, mount the INSTALL media and verify you can load it: # swlist -l fileset -s / | grep Q4 OS-Core.Q4 B.10.10 HP-UX Crash Dump Debugger for PA-RISC systems ^^^^^ This should match OS release! If q4 is not loaded, use swinstall to select and install it: # swinstall -vs / OS-Core.Q4 ============================ STEP 2 ============================ If you wish to direct future dumps to a file system with more space, modify the following line in /etc/rc.config.d/savecore: SAVECORE_DIR=/var/adm/crash <----DEFAULT destination Verify you have a dump to analyze: # ll /var/adm/crash/co* (or your dump directory) You may see: /var/adm/crash/core.0/INDEX /var/adm/crash/core.0/vmunix.gz /var/adm/crash/core.0/core.0.1.gz /var/adm/crash/core.0/core.0.2.gz /var/adm/crash/core.0/core.0.3.gz /var/adm/crash/core.0/core.0.4.gz ^ your suffix may vary The ../core.x/INDEX file and the /etc/shutdownlog contains the "panic" statement. `touch /etc/shutdownlog` if it doesn't exist. If the system dump isn't in the expected location and it hasn't been overwritten by swapping (check swapinfo: % USED=0), try to save the dump with: # savecore -vr ============================ STEP 3 ============================ THE FOLLOWING COMMANDS MUST ALL BE RUN IN THE DUMP DIRECTORY: NOTE: csh (c-shell) will cause errors) 1) # cd to the dump directory eg: cd /var/adm/crash/core.0 NOTE: FOR 10.20/30 DUMPS, SKIP TO STEP 5 2) # uncompress /usr/contrib/lib/Q4Lib.tar.Z (ignore the error if this was done previously) 3) # tar -xf /usr/contrib/lib/Q4Lib.tar (output goes into your current directory) 4) # cp q4lib/sample.q4rc.pl ~/.q4rc.pl (.pl =.p-el) ^ note the tilde (output goes to your login HOME directory) 5) # /usr/contrib/bin/gunzip vmunix.gz (uncompresses the kernel file) 6) If your O/S release is 10.10, type: # /usr/contrib/bin/q4pxdb vmunix (this may complain if vmunix is already preprocessed) If your O/S release is 10.20 or 10.30 type: # /usr/contrib/bin/q4prep -p 7) If in this step, you get the message: "/var: file system full" move the core. directory to a file system with adequate available space and start at this point again. (adequate = approx. 2x the sum of the core.x.y.gz files) Now type: # q4 -p . _____ Note this 'dot' at the end of the line. 8) At the q4> prompt type: trace event 0 > trace If an RCE is on the line type: trace event 0 9) At the prompt type: include analyze.pl (.p-el) 10) At the next prompt type: run Analyze AU >> ana.out NOTE: ctrl-c will interrupt 11) At the next prompt type: exit ============================ STEP 4 ============================ Generate a patch list: # swlist -l product PH\* > patch_list Using the CALL ID as the subject, email the patch_list, ana.out, whathappened.out (if created) and the trace files to: hpcu@hpuerca.atl.hp.com In the future, you may wish expedite a solution by previously performing these steps before logging a response center call. Then email these files to us with the newly assigned callID as the subject and inform the RCE that email is enroute. --------------------------------------------------------------- Latest version of q4: ftp://contrib:9unsupp8@hprc.external.hp.com/crash/q4.1.79.tar.Z *** END *** djm 5/7/96 sed 1/98 FILENAME: /CSCinfo/9000/SWinfo/DownSystems/Dumps/using_q4 LAST MODIFY DATE: Mon Jun 29 09:36:35 EDT 1998