Carl says: >25 minutes later, the process finally dies. >The questions are: >1. What signal killed the process? >2. Why did it take so long? >If what I've said makes sense to you, do you think 25minutes is about >right? I think it is rather too long.
Dumping core in that environment is not qoing to just be (size of core) /(speed of pipe) - there's going to be all sorts of, ah, crud going on. Not inconceivably there will be traffic that goes: memory -> swap -> memory -> core file, when stuff gets swapped out to make room for the process that does the core dump... There's also the bit about you probably not having exclusive use of the pipe, and the "best case" for a 10Mb/s pipe running ethernet being about 5Mb/s... Your numbers went like this, I assume: 780MB core file at 5Mb/s is roughly 15 by 8 bits and some exponent, which is 120 and ... (thinks) ... duh, 1, so 1200 seconds in the best case. So degrading from 20 minutes to 25 seems a surprisingly good result to me. Let me just Feynman my results... 1000 seconds at 5Mb/s is 5000Mb, or 600MB give or take. So the result is at least plausible, if you are in fact using single duplex Ethernet rather than some actual guaranteed 10Mb/s link. Moz _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
