[torqueusers] pbs_server not marking nodes as busy

Garrick Staples garrick at usc.edu
Tue Nov 16 20:24:34 MST 2004


On Fri, Nov 05, 2004 at 02:15:32PM -0800, Garrick Staples alleged:
> On Fri, Nov 05, 2004 at 10:49:53AM -0800, Garrick Staples alleged:
> > Have the semantics around a "busy" node changed?  All of a sudden pbs_server
> > isn't marking nodes that DON'T have a job, but have high load, as busy.
> > 
> > torque-1.1.0p4-snap.1099003850 with all of the recent patches on the list.
> 
> This is caused by restarting the server.  For some reason the BUSY state is
> never sent back to the server.  So either the server needs to request this
> information, or the mom needs to send it?

I think I have a solution here.  If pbs_mom gets an eof from the server, set
the internal_state to update the server.  "It works for me"


diff -ruN torque-1.1.0p5_orig/src/resmom/mom_main.c torque-1.1.0p5/src/resmom/mom_main.c
--- torque-1.1.0p5_orig/src/resmom/mom_main.c	2004-11-16 11:21:20.000000000 -0800
+++ torque-1.1.0p5/src/resmom/mom_main.c	2004-11-16 18:56:21.000000000 -0800
@@ -3301,6 +3305,7 @@
       id,
       dis_emsg[ret]))
 
+    internal_state &= UPDATE_MOM_STATE;
     im_eof(stream,ret);
 
     return;
@@ -3314,6 +3319,7 @@
       id, 
       dis_emsg[ret]))
 
+    internal_state &= UPDATE_MOM_STATE;
     im_eof(stream,ret);
 
     return;

-- 
Garrick Staples, Linux/HPCC Administrator
University of Southern California
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20041116/47901ae3/attachment.bin


More information about the torqueusers mailing list