On Wed, 2004-04-07 at 16:57, Jeremy Warren wrote: > We have been experiencing some weird problems in our DB2/UDB linux guest > and we think that creating memory dumps when the problem is occuring may > help shed some light on the problem (i.e. trying to figure out what sql > queries were executing at the time of the problem).. I have found
The Linux for zSeries support folks can handle dumps, which means setting up a large enough mini disk and prepare that with zipl as your dump volume. You IPL from that device to cause the dump. That is clearly destructive. After that you IPL Linux again you read the dump back in so that yet get a flat file that can be shipped (so you need another chunk of space to hold that file temporarily). That process is described in the document on DeveloperWorks, I think. Using VMDUMP would be more convenient and would not be destructive, but support teams cannot handle that currently. The idea is that we should be able to convert the vmdump output to something lcrash can read again. I have spent some time reading dumps from Linux virtual machines using DVF and VM DumpTool. I can sort of find my way through the table of processes and find memory maps etc, but it's not a lot of fun yet. Whether a dump would give them enough information to analyze the problem is another matter. A lot of stuff is out on disk and swap. Rob ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
