We are also having problems with maui not seeing all the PBS nodes properties. We are using PBSPro 5.4.2.43350 and maui 3.2.6p14. I noticed that whenever I start maui, it sees all the node properties, but on it's second scheduling iteration, it only sees one of the properties. I then found two different parts of the maui code which parses the node information and both parse them differently. So, I went ahead and copied one method of parsing the property information to the other section, rebuilt the code, and now maui sees all the nodes properties. Here's a patch I applied to the MPBSI.c file in the src/moab directory of the maui-3.2.6p14 distribution:

-------------------------------------------------------------------
--- MPBSI.c.orig        2006-01-13 17:24:25.000000000 -0500
+++ MPBSI.c     2006-03-27 09:05:12.000000000 -0500
@@ -3359,19 +3359,20 @@
       }
     else if (!strcmp(AP->name,ATTR_NODE_properties))
       {
-      MUStrCpy(tmpBuffer,AP->value,sizeof(tmpBuffer));
-
-      memset(N->FBM,0,sizeof(N->FBM));
+      if (R->U.PBS.PBS5IsEnabled == FALSE)
+        {
+        strcpy(tmpBuffer,AP->value);

       ptr = MUStrTok(tmpBuffer,", \t",&TokPtr);

       while (ptr != NULL)
         {
-        MUGetMAttr(eFeature,ptr,mAdd,N->FBM,sizeof(N->FBM));
+          MNodeProcessFeature(N,ptr);

         ptr = MUStrTok(NULL,", \t",&TokPtr);
+          } /* END while (ptr != NULL) */
         }
-      }
+      }   /* END else if (!strcmp()) */
     else if (!strcmp(AP->name,ATTR_NODE_np))
       {
       /* set virtual processors */
---------------------------------------------------------

Use this at your own risk! My experience with the maui support team is that everything is going into moab and if you want a REAL system, you'll purchase moab. Can't blame them for making a buck, but I figured I'd just look into the maui source code to find these problems and fix them myself.

And note, I'm not a developer - I'm a system admin with some programming skills. So no comments about my patch file. I do my best.

Jenett

--
  Jenett Tillotson, System Administrator
  Rosen Center for Advanced Computing, Purdue University
  [EMAIL PROTECTED] / v: (765) 494-9387


----------------------------------------------------------------------

Message: 1
Date: Wed, 22 Mar 2006 16:10:56 -0600
From: "Franco M. Bladilo" <[EMAIL PROTECTED]>
Subject: [Mauiusers] maui not honoring multiple PBSPro node attributes
To: [email protected]
Message-ID: <[EMAIL PROTECTED]>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

We are having issues with maui-3.2.6p14 and multiple PBSpro node
attributes, maui doesn't seem to see after the first node attribute on a
pbs node specification , for example:

ada759-6:/var/spool/PBS/server_priv # cat nodes | grep matlab
ada746-6 pcpus=4 fpga matlab amd275
ada746-5 pcpus=4 fpga matlab amd275
ada748-1 pcpus=4 amd275 matlab
ada748-3 pcpus=4 amd275 matlab
ada748-2 pcpus=4 amd275 matlab
ada748-5 pcpus=4 amd275 matlab
ada748-6 pcpus=4 amd275 matlab
ada748-4 pcpus=4 amd275 matlab
ada759-6:/var/spool/PBS/server_priv # diagnose -n | grep ada74
ada746-1                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [fpga]
ada746-2                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [fpga]
ada746-3                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [fpga]
ada746-4                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [fpga]
ada746-6                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [fpga]
ada746-5                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [fpga]
ada748-1                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [amd275]
ada748-3                 Idle   4:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 000
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [amd275]
ada748-2                 Busy   0:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 002
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [amd275]
ada748-5                 Busy   0:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 001
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [amd275]
WARNING:  node 'ada748-5' is active for 00:41:34 but load is LOW.
load:  0.000 (check job 44262?)
ada748-6                 Busy   0:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 001
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [amd275]
WARNING:  node 'ada748-6' is active for 00:41:34 but load is LOW.
load:  0.000 (check job 44262?)
ada748-4                 Busy   0:4     8016:8016        1:1
8016:8016    1.00 DEFAUL  linux DEF   0.00 001
[interactive_4:4][phyz_4:4][co [DEFAULT]                      [amd275]
WARNING:  node 'ada748-4' is active for 00:41:34 but load is LOW.
load:  0.000 (check job 44262?)

The node features column only sees the first attribute specified on the
nodesfile , nothing else. Jobs that request the second attribute never
get to run.

Any ideas?

Thanks in advance,


_______________________________________________
mauiusers mailing list
[email protected]
http://www.supercluster.org/mailman/listinfo/mauiusers

Reply via email to