Adding additional worker nodes causes job failure, collections don't error out

[MoHeydarian]

When jobs/collections are running and additional worker nodes are added to the Cloudman instance, some of the jobs/collections stop running and return empty jobs/collections in the that are green state. Not all collections exhibit this behavior, some collections are stopped and in the error state.

The two issues:

• Adding workers interrupts ongoing jobs
• Interrupted jobs are not always marked in the error state, collections contain empty data items

Behavior observed using the GVL 4.4.0 RC2 (Galaxy 18.05)

[MoHeydarian]

Update: removing idle worker nodes while jobs are running leads to the same issue above.

[almahmoud]

Should be fixed with the new CloudMan image, merged here: #90 and deployed to bucket by Enis