Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

Users submitting jobs to their local Institutional Pool can expand their resources by overflowing their jobs to the Global Pool via CMS Connect. This mean user jobs will start running on the local pool and idle jobs still waiting to run will communicate with CMS Connect so that they can be matched with available CMS resources in the Global Pool. In order to do this, the user needs to:

  • Before starting, make sure your system administrator has properly configured condor to allow authentication between the local Institutional Pool and Global Pool resources (see the admin guide).
  • Run a server on the Institutional Pool submit host to communicate with CMS Connect.
    • You will need an open port for this too, ask (default port is: 11001, but you can choose any available). Ask your administrator if you don't know any of this.
  • Login to login.ucms.org and run a client there that can the client. This will read the jobs from the server in their Institutional Pool.

Also, the Institutional pool needs to be properly configures configured (see the admin guide) so that authentication works properly.

...

Code Block
languagebash
titlepyglidein
$ cd pyglidein-master
[pyglidein-master]$ ./server.py -p 50000 -u <you_username>

...

<your_username>

To verify the chosen port is open, you can login into CMS Connect and run nmap to that specific port. The State alternatives are:

  • Open:  This means the server port is opened and listening. 
  • Closed: The server is not running
  • Filtered: The server port is behind a firewall

Code Block
languagebash
titlenmap
[login.uscms.org]$ nmap your_institutional_submit_host.com -Pn -p 50000
Starting Nmap 5.51 ( http://nmap.org ) at 2016-09-15 09:47 CDT
Nmap scan report for opteron03.crc.nd.edu (129.74.85.68)
Host is up (0.012s latency).
PORT      STATE SERVICE
50000/tcp open  ibm-db2
Nmap done: 1 IP address (1 host up) scanned in 0.07 seconds

The client side

Follow the same steps to download the pyglidein tool on login.uscms.org and run it as followed:

Code Block
languagebash
titlepyglidein
[pyglidein-master]$ ./client.py --conf=you_config

There are different configuration examples in the github repository. Use this example as a guide and contact the support team for further help.


References

[1] http://research.cs.wisc.edu/htcondor/HTCondorWeek2016/presentations/ThuSchultz_IceCubeGlideins.pdf

Support

Please, using the following e-mail for any questions: cms-connect-support@cern.ch