Hi guys, I took a look around the archives here, but haven't seen anyone else with this problem - I'm using the 2.5.5 source, and have compiled gmond for use on some diskless Opteron nodes. When I run the gmond executable on the head node, it works fine, but when I try to run it on the diskless nodes, I get a segfault, and the following in my syslog:
kernel: gmond[839]: segfault at 0000000000000000 rip 0000000000000000 rsp 00000000407ff7f8 error 14 .. Obviously I'm inclined to think it's the kernel - so more about the diskless nodes. They ARE compiled with multicast support, and it seems to me that the crash has something to do with the threads (but it's late here, so I haven't gone nuts looking exactly where yet!). I'll probably test booting the kernel as a 'real' (ie, non-diskless) config tomorrow when I have physical access to the machine, but all else seems to work fine. If I mount /usr from my head node, and then run gdb on the gmond executable, it stops with : "Cannot find new threads: generic error". (Also, the root file system is in a ramdisk, not over NFS) Any advice? Thanks, - Brian
