If an application is installed to a SAN and the process crashes, can a core dump be written to the SAN partition? I've been trying to track down an issue where our process dies and no core is generated, thus no stack. Both cases were in two completely separate Solaris environments. ulimit -c is unlimited, pstack is in the path. In my initial research I found a statement claiming it is not possible to core dump to a SAN disk. I have been unable to find this information again, leading me to question its validity.
A:
The conditions when no core is generated or listed in core(4), use man core to read the manpage. If the user id nobody is involved, that might happen (or rather, nothing is happening).
You can also use gcore to see if manually obtaining a core dump will help. Check with coreadm, where core dumps go by default -- maybe they are not in the current working directory, but somewhere in /var?
Volker
ShiDoiSi
2009-02-18 09:56:59
A:
The reference you found to core dumps to a SAN not working is probably referring to core dumps of the kernel itself. (usually called crash dumps on Solaris)
Perhaps the process does not have a useful current directory. Use coreadm to force core dumps to be written somewhere reasonable. See http://developers.sun.com/solaris/articles/manage_core_dump.html
If your process is dying unexpectedly and you have the luxury of destroying its performance, try running it under truss. If this is Solaris 10, dtrace could be quite helpful here.
chuck
2009-05-22 19:52:48