Jump to content

Usage of HPC and HA at BITS Pilani

From TetraWiki
Revision as of 14:12, 29 December 2012 by Biswajit (talk | contribs) (Created page with "<center>'''UGAGE DOCUMENT OF HPC CLUSTER AND HIGH AVAILABILITY CLUSTER'''</center> <center>'''NAVDEEP MATHUR'''</center> <center>'''HIGH AVAILABILITY CLUSTER'''</center> ...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
UGAGE DOCUMENT OF HPC CLUSTER AND HIGH AVAILABILITY CLUSTER


NAVDEEP MATHUR


HIGH AVAILABILITY CLUSTER

Two Nodes on HA:

  1. ibm-39.bits.ac – 172.20.4.39
  2. ibm-40.bits.ac – 172.20.4.40

Virtual IP of HA : 172.20.4.43

How to Fence from one cluster to another manually , Run command from any node

# clusvcadm –d nfs

Here nfs is an cluster service

# clusvcadm –e nfs –m `hostname`

How to check status of cluster

# clustat

[[Image:]]

Stop manually HA Cluster by:

# service cman stop //Simultaneously as root on ibm-39 and ibm-40

# service rgmanager stop //Simultaneously as root on ibm-39 and ibm-40

Start manually HA Cluster by:

# service cman start //Simultaneously as root on ibm-39 and ibm-40

# service rgmanager start //Simultaneously as root on ibm-39 and ibm-40

Then check the status of cluster by “clustat” command and also ping to 172.20.4.43 that is your virtual ip of cluster.

Note :

  1. When first server is out of the network then it fenced to second server and at same time first server has automatically forcefully rebooted
  2. Then virtual ip switch to second server.
  3. When first server has been UP then need to switch cluster manually.
  4. You have two choices by clusvcadm or restart rgmanager and cman.
  5. We take here second option i.e. restart rgmanager and cman

On Node2 :

# service rgmanager stop

# service cman stop

Then Simultaneously on Node1 and Node2

# service cman start

# service rgmanager start

And check clustat


HIGH PROCESSING CLUSTER

When System restart run command as wolf and lam user

For HPC1 -2 node HPC

Login as wolf

#mpdboot –n 2 –f mpd.hosts

#mpdtrace // check active nodes

This will active mpd deamon on all the nodes

Login as lam

#mpdboot –n 2 –f mpd.hosts

#mpdtrace // check active nodes

This will active mpd deamon on all the nodes


For HPC2 -9 node HPC

Login as wolf2

#mpdboot –n 9 –f mpd.hosts

#mpdtrace // check active nodes

This will active mpd deamon on all the nodes

Login as lam1

#mpdboot –n 9 –f mpd.hosts

#mpdtrace //check active nodes

This will active mpd deamon on all the nodes

Then mpi commands execute using mpiexec command