       Condor V6.0 Checkpoint Server Installation Instructions

1. Install Condor on your checkpoint server machine.  (See NOTE #1
below.)

2. Untar the release.tar file into your release directory (i.e., the
directory that contains the sbin, bin, lib, and etc directories for
your main Condor installation).  This will create the following files:

    sbin/condor_ckpt_server
    sbin/condor_cleanckpts
    etc/examples/condor_config.local.ckpt.server

3. Create a directory where the checkpoint server will store the
checkpoint files.

4. Replace the local config file for the checkpoint server machine
with the condor_config.local.ckpt.server file provided.  You will need
to customize this file to specify the directory used for storing
checkpoint files.

5. Set the following parameters in your pool-wide config file:

	USE_CKPT_SERVER = True
	CKPT_SERVER_HOST = condor-ckpt.at.your.domain

replacing "condor-ckpt.at.your.domain" with the hostname of your
checkpoint server machine.

NOTES:

1. Your checkpoint server machine does not require a full Condor
install, although this is the most convenient method to set up the
machine.  Only the following components are actually required:
	- condor_config file
	- MASTER binary
	- LOG directory

2. In some cases, old checkpoint files may be left on the checkpoint
server.  The condor_cleanckpts script is provided to remove these
files.  This script requires that condor_config_val and condor_q are
correctly installed on your checkpoint server.  You may want to run
this script periodically via cron(1).

3. All of the machines in your pool do not need to use the same
checkpoint server.  You may configure multiple checkpoint server
machines.  Set CKPT_SERVER_HOST in the local config file of each
machine in your pool to the hostname of the checkpoint server to be
used for jobs submitted from that machine.  Each machine may not use
more than one checkpoint server, however.  We hope to overcome this
limitation soon.

