From: Paul H. Hargrove (PHHargrove_at_lbl_dot_gov)
Date: Wed Apr 23 2008 - 12:18:44 PDT
I believe your problem is probably "prelinking", as described in the following FAQ entry: http://mantis.lbl.gov/blcr/doc/html/FAQ.html#prelink Let us know if after disabling prelinking you still have problems moving checkpoints between nodes. -Paul ǿ �� wrote: > My cluster has 64 nodes with Linux 2.6.20, x86-64 . > MPI programs make checkpoint with BLCR-0.5.5, ans restarted successfully > on the same node, e.g. cn01. But programs receive signal SIGSEGV in > libc when restarted on the nodes other than cn01. > I find /lib64/tls/libc.so.6 has different entry point on different nodes. > How to do? > > ------------------------------------------------------------------------ > �Ż����䣬�����������䣡 <http://cn.mail.yahoo.com/> -- Paul H. Hargrove PHHargrove_at_lbl_dot_gov Future Technologies Group HPC Research Department Tel: +1-510-495-2352 Lawrence Berkeley National Laboratory Fax: +1-510-486-6900