The pantasks "server" mode allows to pantasks to run using a server/client interface. This is especially handy if you want to be able to control pantasks from multiple locations (e.g., check processing from home) or accounts (e.g., multiple people controlling or eavesdropping on the processing).
There is a small convenience issue when running pantasks in the "server" mode: some commands which you would usually run in a standalone pantasks are only available when wrapped inside a macro; see below.
=== Configuration ===
Ensure the following is in your ~/.ptolemyrc file:
PANTASKS_SERVER XXXXX.ifa.hawaii.edu
PASSWORD XXXXX
#PANTASKS_SERVER_STDOUT pantasks.stdout.log
#PANTASKS_SERVER_STDERR pantasks.stderr.log
* PANTASKS_SERVER defines the name of the server for the client to connect to. There's no real need to update this, since we'll use a script to set this on the fly, allowing multiple pantasks servers to be used simultaneously.
* PASSWORD provides minimal access control for the pantasks server. Set this to a unique word.
* PANTASKS_SERVER_STDOUT and PANTASKS_SERVER_STDERR define where the output goes that is usually displayed in the pantasks session. Since setting these also affects the standalone version of pantasks, I recommend leaving them commented out; we'll use a script to set them on the fly.
If you are using Nebulous, you will also need to set up an ipphosts.config file. This config file has METADATA cards that assign a particular host to each chip on the camera, so that nebulous can distribute every exposure consistently, without piling too many chips onto the same machine. There is also a metadata card for indicating which hosts are to be used for skycell processing. Here is an example of how this config file should look:
ipphosts MULTI
ipphosts METADATA
camera STR skycell
count S32 8
sky00 STR ipp006
sky01 STR ipp007
sky02 STR ipp008
sky03 STR ipp009
sky04 STR ipp010
sky05 STR ipp011
sky06 STR ipp020
sky07 STR ipp021
END
ipphosts METADATA
camera STR GPC1
XY01 STR ipp014
XY02 STR ipp014
XY03 STR ipp014
XY04 STR ipp014
XY05 STR ipp023
XY06 STR ipp023
XY10 STR ipp023
XY11 STR ipp023
XY12 STR ipp024
XY13 STR ipp024
< etc. ... >
XY75 STR ipp015
XY76 STR ipp015
END
=== Helpful scripts ===
Here are some convenient scripts for starting the pantasks server and client.
* Save the following script as '''start_pantasks_server''' in some directory on your path
#!/bin/bash
# Start up a pantasks server
# Output is redirected to files
server_redirect="-D PANTASKS_SERVER_STDOUT pantasks.stdout.log -D PANTASKS_SERVER_STDERR pantasks.stderr.log"
echo pantasks_server $server_redirect
pantasks_server $server_redirect
* Save this one as '''start_pantasks_client''' on your path
#!/bin/bash
# start up pantasks_client
# 1st argument is the name of the host to connect to
server_arg=""
server_host=$1
if [ $server_host ]; then
server_arg="-D PANTASKS_SERVER $server_host"
echo pantasks_client $server_arg
pantasks_client $server_arg
else
echo "Please provide the hostname of the pantasks server."
fi
=== Starting pantasks ===
Starting the server is as easy as:
% start_pantasks_server
bound to port: 2000
You can start it in the background if you want: it's not going to print anything more to the terminal.
In a different window (or machine or user), start the client:
% start_pantasks_client mymachine
where mymachine is the name of the machine on which the server is running.
You should be presented with the usual pantasks environment. However, some things are subtly different: the environment you see before you is particular to the client, not the server, and it's the server that runs everything. This is important to keep in mind, and it means that you should bury within macros some commands that you would ordinarily just type on the pantasks command-line.
=== Setting up ===
Here are some useful definitions for controlling pantasks using the server mode:
* Save this as server.pro in your pantasks modules directory
macro setup
$LABEL = "my_label"
$DBNAME = "mydatabase"
module pantasks.pro
module.tasks
all.off
# NEXT LINES ARE ONLY NEEDED IF USING NEBULOUS
$default_host = machine1
queueload tmp -x "cat ipphosts.config"
ipptool2book tmp ipphosts -key camera
# END NEBULOUS ONLY
add.database $DBNAME
echo Using database $DBNAME
echo Current active label is $LABEL
end
macro change_label
$LABEL = $1
echo Current active label is $LABEL
end
macro print_label
echo Current active label is $LABEL
end
macro machines
controller host add machine1 -threads $1
controller host add machine2 -threads $1
controller host add machine3 -threads $1
controller host add machine4 -threads $1
end
* If you're not using Nebulous, you can remove the lines that are specific to Nebulous use (these send jobs to particular machines).
* Ensure the my_label is changed to the label of the data you're processing (or "NONE" including quotes if none).
* Ensure the my_database is changed to the database you're using.
* Ensure the machineN are changed according to your system.
So, now in the pantasks client, you can say:
pantasks: server input server.pro
to load the macro file. Note that the server has to do this command, hence the prefix server.
Here is how you would set up for processing, and add hosts with a maximum of 4 threads each:
pantasks: setup
pantasks: machines 4
Now you should be able to simply say:
pantasks: run
and pantasks will run as usual. The status, stop and halt commands are also available as usual (no need to prepend server or anything like that). But if you want to change the label, then you should do:
pantasks: change_label "new_label"
whereas in the standalone pantasks, you could have done:
pantasks: $LABEL = "new_label"
Such commands now have to be buried within macros.
=== Shutdown ===
You can quit out of the client at any time by saying:
pantasks: quit
This does not stop the server from continuing to process. You can reconnect to the server as you please.
When you're all done, you can force the server to quit by issuing the following command from the client:
pantasks: shutdown now