[SCore-users-jp] Re: [SCore-users] Problem starting multi-user
Atsushi HORI
hori @ swimmy-soft.com
2002年 4月 15日 (月) 18:41:12 JST
Hi.
>Hi we had a problem with one of our cluster nodes (comp3) so commented it out
>of scorehosts.db.
>
>I have restarted scoreboard and msgbserv.
>
>Now when starting scored with sc_watch I get the following error:
SCore-D remebers the cluster configuration of the last time and also
remembers the jobs running. Here "remeber" means some info is stored
into disk. When SCore-D is restarted, then it checks the file with
the current cluster configuration, and when it discovers some
difference, then it will not to start.
When you want to start SCore-D with different clsuter ocnfiguration,
you must specify "session No." or reset the previous session with
"-reset" option.
The session was introduced so that aborted job(s) in the previous
session can be preserved with the different configuration. Think
about the case two hosts crash while some user jobs are running. The
jobs only restarted from their checkpoint when the two hosts are
fixed, but it may takes some time. And you want to continue SCore-D
running on the shrinked cluster (beacuse two hosts are missing).
This can be done by specifying the different session id. When the two
hosts are fixed, then stop SCore-D and restart SCore-D with the
original session no.
The default session no. is 0.
_______________________________________________
SCore-users mailing list
SCore-users @ pccluster.org
http://www.pccluster.org/mailman/listinfo/score-users
SCore-users-jp メーリングリストの案内