[torqueusers] Torque 1.1.0p3

scott at supercluster.org scott at supercluster.org
Mon Oct 18 12:19:16 MDT 2004


Torque Users,

Torque 1.1.0 patch 3 is now available.  It contains two major improvements and 
many minor enhancements.

#1  momctl

   The momctl command allows diagnosis and management of the mom daemon and uses 
the following syntax:

USAGE:  momctl <ARGS>
             [ -c [JOB] ]       // CLEAR STALE JOB
             [ -d DIAGLEVEL ]   // DIAGNOSE (0 - 3)
 	    [ -f HOSTFILE ]    // HOSTFILE
             [ -h [HOST] ]      // HOST
             [ -p [PORT] ]      // PORT
             [ -q [ATTR] ]      // QUERY STATE
             [ -r [FILE] ]      // RECONFIG
             [ -s ]             // SHUTDOWN

   If diagnostics are requested, the following information is reported:

-----
Host: hana/hana.icluster.org  Server: 10.10.10.106
HomeDirectory:         /usr/spool/PBS/mom_priv
MOM active:            240 seconds
Last Msg From Server:  240 seconds (CLUSTER_ADDRS)
Last Msg To Server:    6 seconds
Init Msgs Received:    0 hellos/1 cluster-addrs
Init Msgs Sent:        1 hellos
LOGLEVEL:              0 (use SIGUSR1/SIGUSR2 to adjust)
Trusted Client List: 
10.10.10.110,10.10.10.121,10.10.10.119,10.10.10.117,10.10.10.106,127.0.0.1
JobList:               NONE
-----

   In addition, a number of internal checks are performed and failures reported.

   If there is additional information which sites may find of value in 
diagnosing local failures, please let us know.


# 2  Modification of the pbs_mom to pbs_server daemon initialization sequence

   Using momctl, we finally were able to isolate and correct an issue which has 
been plagueing a number of sites.  There was an issue with the way moms 
initialized communication with the pbs_server resulting in 'unexpected eof' and 
'cannot connect' messages.  This change should not only remove these failures, 
but it should also accelerate pbs recycle times allowing all nodes to come on 
line faster.

   Please test patch 3 and let us know of any issues you see.  Thanks for all of 
the contributed patches.

Supercluster Development Group


More information about the torqueusers mailing list