Removing or replacing a node

First see the section for how to remove or add the hardware safely.

If the removed compute node is a multi-user node you need to comment it out of the configuration file

/opt/score/etc/scorehosts.db

on the front end server and restart the scoreboard daemon:

[root@scserve streamline]# /etc/rc.d/init.d/scoreboard stop
Shutting down scoreboard services:                         [  OK  ]
[root@scserve streamline]# /etc/rc.d/init.d/scoreboard start
Starting scoreboard services:                              [  OK  ]

NOTE :- the file scorehosts.db is a set of C preprocessor directives NOT a shell script so you should comment out lines as you would in C using

 
/* This is a comment */

To replace a node simply work the above procedure in reverse.

If the removed node is a batch node (PBS or SGE) then you do not need to comment it out, as the batch scheduler will recognise automatically that the compute node is unavailable.

NOTE : - If you are given a new replacement compute node then it will have a hardware ethernet address different from the one that it is replacing. You should then edit the file /opt/score/etc/pm-ethernet.conf which gives a hostname - MACH address mapping and also the file whose name is the number of the new node in /opt/score/etc/ndconf - e.g /opt/score/etc/ndconf/32 for node 32.