Normal run

• the idea is to just kill the LoggingManager (which is the thing that gives out the events to the elven), so when the elven do not get any more events to process they just finish nicely.
• you can kill the LM by just killing the process on the oprserver, but a nicer way to do that is with the nameserver.
• the nameserver runs on the oprserv machine and helps the LM communicate with the elven running on the nodes.
  The LM registers itself with the nameserver and the nodes look it up using the NS.
• you can effectively 'unregister' the LM with the NS which causes the process to die.

Here's the recipe:

1. need to set the IOR file for the nameserver you want to look at:
   e.g. for PC1 (alias nameserver1)
   setenv NameServiceIOR `cat /nfs/oprserv01/u1/TAO/TAO_NS.ior`
2. make sure you have done srtpath (newish release)
3. use TaoNSDumper to look at the nameserver
   > TaoNSDumper

   Name service ior:
   IOR:010000002b00000049444c3a6f6d672e6f72672f436f73 ... 
   
   Name Graph
        0:  PC1: context
              LM: context

   here no LM is registered.
   With a registered LM you will see:
   > TaoNSDumper

     Name service ior:
   IOR:010000002b00000049444c3a6f6d672e6f72672f436f734 ... 
   
   Name Graph
        0:  ER6: context
          0:  LM: context
            0:  44279-1: reference

4. use OprLMInterrupter to kill the LM
   > OprLMInterrupter ER6/LM/44279-1
5. You will see that the elven start to finish and slowly (order 5min) the run should finish.

Other Ways to Kill a Run

RP not started

-> can reap and scratch
check first that the run has not started:
- look for the logdir
- look in oprruns status=Processing
- check no elven running:
findElven ER1

otherwise you want the farm to be asleep
before reaping, and reaped before scratching,
except in emergency cases.

If the RP is running... want to put it to sleep.

- there is a logdir
- there is oprruns status=Processing

- look in logdir to see what is happening
- look for running LM
- look for elven: findElven ER1

to kill nicely:
> OprLMInterrupter ERx/LM/run#-1

to kill meanly:
killElven ER1

if all elven are dead or not yet started, and
you know the processing is bad:

rpsetstate ER1 SetFailedProcessing (only for ER farms)

if do this while elven or LM are running they
will not be killed, run will continue, but CS will
not monitor it anymore.
if RP is stuck in MonitorProcessing (i.e. some NPs do not return) you can do:
rpsetstate ER# RecoverProcessing(works on ER and PC farms)
This will allow the run to finish, it will automatically get marked done or failed as appropriate.

Run with Nodes in bad shape (with never started elvens)