'Unknown' status when performing batch analysis in ColumbusPlus

'Unknown' status when performing batch analysis in ColumbusPlus

The job essentially gets stuck in this state and does not proceed. It is caused by a failure in retrieving the job state typically caused by an ssh timeout. A retry strategy has been added to future versions of ColumbusPlus (2.9 onward). However in earlier versions ssh can be configured to retry. ssh can also be configured to make the connection persistent.

Retry

Create the directory ~columbus/.ssh on the Columbus server, accessible only to the Columbus user.

Create a file named config in the above path which contains the single line:

  ConnectionAttempts 3

ssh will only stop trying after experiencing the third timeout.

Persistent Connection

In the file ~columbus/.ssh/config add additional lines:

  ControlMaster auto

  ControlPath ~/.ssh/socket-%h-%r

  ControlPersist 300

This will create a persistent ssh connection to the cluster, listening at the socket in ~/.ssh. From thence forward, all ssh requests will go through the above socket as oppose to opening new connections. The connection will be closed if left idle for more than 5 minutes, but the next ssh request will automatically recreate the session.