mnodectl

Synopsis

       mnodectl -d message:<index> nodeexp       mnodectl -m       attr{=|-=}val       nodeexp       mnodectl -q [cat|diag|profile|wiki] nodeexp

Overview

Change specified attributes for a given node expression.

Access

By default, this command can be run by any Moab Administrator.

Format

-d - Destroy
message:<INDEX>
N/A

With message:<INDEX> argument: Remove the message with the given index from the node.

> mnodectl node001 -d message:admin2
The admin2 message on node001 is removed.
  
-m - Modify

<ATTR>{=|-=}<VAL>

Where <ATTR> is one of the following:
GEVENT,
GMETRIC,
MESSAGE,
OS,
POWER,
VARIABLE

and -=, in this case, clears the attribute; it does not decrement the attribute's value.

NoteChanging OS and POWER require a Moab Adaptive Computing Suite license and a provisioning resource manager.
---
Modify the state or attribute of specified node(s)
> mnodectl -m gevent=cpufail:'cpu02 has failed w/ec:0317' node1 
> mnodectl -m gmetric=temp:131.2 node1 
> mnodectl -m message='cpufailure:cpu02 has failed w/ec:0317' node1 
> mnodectl -m OS=RHAS30 node1 
> mnodectl -m power=off node1
> mnodectl -m variable=IP=10.10.10.100,Location=R1S2 node1 
  
-q - Query
{cat | diag | profile | wiki}
---

Query node categories or node profile information (see ENABLEPROFILING for nodes).

NoteThe diag and profile options must use --xml.
> mnodectl -q cat ALL
node categorization stats from Mon Jul 10 00:00:00 to Mon Jul 10 15:30:00
Node: moab
 Categories:
                     busy: 96.88%
                     idle: 3.12%
Node: maka
 Categories:
                     busy: 96.88%
                     idle: 3.12%
Node: pau
 Categories:
                     busy: 96.88%
                     idle: 3.12%
Node: maowu
 Categories:
                     busy: 96.88%
                  down-hw: 3.12%
Cluster Summary:
                     busy: 96.88%
                  down-hw: 0.78%
                     idle: 2.34%
> mnodectl -v -q profile
...
> mnodectl -q wiki <nodeName>
        
Query a node with the output displayed in a WIKI string. The node's name may be replaced with ALL.

Parameters

GEVENT
<EVENT>:<MESSAGE>
---
Creates a generic event on the node to which Moab may respond. (see Enabling Generic Events)
mnodectl -m gevent=powerfail:'power has failed' node1
  
GMETRIC
<ATTR>:<VALUE>
---
Sets the value for a generic metric on the node. (see Enabling Generic Metrics).
NoteWhen a gmetric set in Moab conflicts with what the resource manager reports, Moab uses the set gmetric until the next time the resource manager reports a different number.
mnodectl -m gmetric=temp:120 node1
  
MESSAGE
'<MESSAGE>'
---
Sets a message to be displayed on the node.
mnodectl -m message='powerfailure: power has failed' node1
  
NODEEXP
<STRING>
Where <NODEEXP> is a node name, regex or "ALL"
---
Identifies one or more nodes.
node1 - applies only to node1
fr10n* - all nodes starting with "fr10n"
ALL - all known nodes
  
OS
<STRING>
---
Operating System (see Resource Provisioning)
mnodectl node1 -m OS=RHELAS30
  
POWER
{off|on}
---

Set the power state of a node. Action will NOT be taken if the node is already in the specified state.

Note: If you power off a node, a green policy will try to turn it back on. If you want the node to remain powered off, you must associate a reservation with it.

Note

If you request to power off a node that has active work on it, Moab will return a status indicating that the node is busy (with a job or VM) and will not be powered off. You will see one of these messages:

  • Ignoring node <name>: power ON in process (indicates node is currently powering on)
  • Ignoring node <name>: power OFF in process (indicates node is currently powering off)
  • Ignoring node <name>: has active VMs running (indicates the node is currently running active VMs)
  • Ignoring node <name>: has active jobs running (indicates the node is currently running active jobs)

Once you resolve the activity on the node (by preempting or migrating the jobs or VMs, for example), you can attempt to power the node off again.

You can use the --flags=force option to cause a force override. However, doing this will power off the node regardless of whether or not its jobs get migrated or preempted (i.e., you run the risk of losing the VMs/jobs entirely). For example:

> mnodectl node1 -m power=off --flags=force
> mnodectl node1 -m power=off
  
VARIABLE
<name>[=<value>],<name[=<value]]...
---
Set a list of variables for a node.
> mnodectl node1 -m variable=IP=10.10.10.100,Location=R1S2
      

See Also

Copyright © 2012 Adaptive Computing Enterprises, Inc.®