Dying client?

[bill]

I think I have a cluster working but my clients keep dying. At 1:05 one job died on a node, at 1:20 another died.
If I look at the dir:




node001:~/md5crk/node001/a> ls -alt
total 60
-rw-r--r-- 1 bill users 991 Jul 6 13:18 md5_perf.xml
-rw-r--r-- 1 bill users 12673 Jul 6 13:18 md5_send.xml
-rw-r--r-- 1 bill users 1050 Jul 6 13:10 md5_save.xml
drwxr-xr-x 3 bill users 4096 Jul 6 11:55 .
bill@node001:~/md5crk/node001/b> ls -alt
total 52
-rw-r--r-- 1 bill users 991 Jul 6 13:05 md5_perf.xml
-rw-r--r-- 1 bill users 12487 Jul 6 13:05 md5_send.xml
-rw-r--r-- 1 bill users 1050 Jul 6 12:47 md5_save.xml
drwxr-xr-x 3 bill users 4096 Jul 6 11:55 .


In neither case is md5_gui.log helpful or current. In neither case is CurrentDPCount at 200. In neither case is the daemon still running. Any ideas?

Oh, if I'm using md5_preset without a proxy do I skip the <proxy></proxy> line or just leave it
empty?

[bill]

I think even (under linux) when I run md5crk -bg that it keeps either
stderr or stdout open. I suspect that if I kill that session that once the buffer
gets full that the client died. Possible? I've noticed if I ssh nodexx md5crk -bg
I can't close that session (even after getting the prompt back) unless I control-c.

[jlcooke]

-gb seems to work on my box. I close stdout/err/in but perhaps I missed something.

When I background other tasks in linux and logout via ssh and the ssh shell still holds on, I need to close the ssh.

Ctrl-C may not do this. Enter ~ . is the standard escape sequence for ssh. Try that while I look into fixing this properly. Thanks for being patiet, I want to make this as easy as possible for people like you so your input is very much apariciated.