GEMSupervisor stuck during state transitions
Description
There are a few conditions in which the GEMSupervisor
can get stuck during state transitions. It can be due to (1) an undefined global state or (2) unknown reasons such as in http://cmsonline.cern.ch/cms-elog/1169110.
While condition (1) should not happen in regular operations, condition (2) can. This should be understood and fixed.
Steps to reproduce
For (1), it is enough to put the system in an undefined state (e.g. triggering state transitions directly in the AMCManager
) before triggering state transitions from the GEMSupervisor
. For (2), it is a rare and difficult situation to reproduce as stated in the e-log. So far, I haven't been able to reproduce it.
Possible fixes
Re-write the GEMSupervisor
state transitions to make them robust.