[Swift-devel] Using multicore servers as Swift pools

[Michael Wilde]

Luiz,

An example of the config files you will need in order to use the 10 8-core 64-but MCS compute servers as Swift pools is on the CI net under /home/wilde/swift/lab/{coasters.xml,auth.defaults.sample}

The servers (10 x 64 bit and 3 x 32 bit) are listed at:
  http://wiki.mcs.anl.gov/IT/index.php/General_MCS_Questions#computeservers

Since these machines are behind a firewall, you can use the ~/.ssh/config example below (adapted as needed) to make them accessible as if they permitted direct login. You login to login.mcs.anl.gov using your local ssh key, and then ssh ports are forwarded to each of the target machines that you want on the MCS network.  The technique is explained at:
  http://articles.techrepublic.com.com/5100-10878_11-6155832.html

This is one of the configurations we should test with new test scripts for Swift 0.91

Ive pasted the files below as well.

- Mike

=== auth.defaults.sample ===

xlogin1.pads.ci.uchicago.edu.type=password
xlogin1.pads.ci.uchicago.edu.username=wilde

login.pads.ci.uchicago.edu.type=key
login.pads.ci.uchicago.edu.username=wilde
login.pads.ci.uchicago.edu.key=/home/wilde/.ssh/swift_rsa
login.pads.ci.uchicago.edu.passphrase=mypassphrasegoeshere

login.mcs.anl.gov.type=key
login.mcs.anl.gov.username=wilde
login.mcs.anl.gov.key=/home/wilde/.ssh/swift_rsa
login.mcs.anl.gov.passphrase=mypassphrasegoeshere

=== ~/.ssh/config set up for forwarding one compute server, "crush"

Host * 
  ServerAliveInterval 15

#ControlMaster auto

#ControlPath ~/.ssh/ssh-connections/%r@%h:%p

#  COMPUTEHOSTS='crush thwomp stomp crank steamroller grind churn trounce thrash vanquish'

Host mcs login.mcs.anl.gov
  Hostname login.mcs.anl.gov
  ForwardAgent yes
  ForwardX11 no
  LocalForward 19001 140.221.8.62:22
 
Host crush thwomp stomp crank steamroller grind churn trounce thrash vanquish
  ForwardAgent yes
  ForwardX11 no
  Hostname localhost
  NoHostAuthenticationForLocalhost yes
 
Host crush
  Port 19001
 
=== coasters.xml ===

<config>
  <pool handle="crush">

    <execution provider="coaster" url="crush.mcs.anl.gov" jobmanager="ssh:local"/>

    <profile namespace="globus" key="workersPerNode">8</profile>
    <profile namespace="globus" key="maxTime">3500</profile>
    <profile namespace="globus" key="slots">1</profile>
    <profile namespace="globus" key="nodeGranularity">1</profile>
    <profile namespace="globus" key="maxNodes">1</profile>

    <filesystem provider="local" url="none" />
    <profile key="jobThrottle" namespace="karajan">.07</profile>
    <profile namespace="karajan" key="initialScore">10000</profile>
    <workdirectory>/home/wilde/swiftwork/crush</workdirectory>
  </pool>

  <pool handle="thwomp">
    <execution provider="coaster" url="thwomp.mcs.anl.gov" jobmanager="ssh:local"/>
    <profile namespace="globus" key="workersPerNode">8</profile>
    <profile namespace="globus" key="maxTime">3500</profile>
    <profile namespace="globus" key="slots">1</profile>
    <profile namespace="globus" key="nodeGranularity">1</profile>
    <profile namespace="globus" key="maxNodes">1</profile>
    <filesystem provider="local" url="none" />
    <profile key="jobThrottle" namespace="karajan">.31</profile>
    <profile namespace="karajan" key="initialScore">10000</profile>
    <workdirectory>/home/wilde/swiftwork/thwomp</workdirectory>
  </pool>

...etc for the rest of the 10 compute servers...

- Mike