Page History

...

If you see jobs being terminated with exceptions in the message viewer saying things like "yili0148+5: Host or host group is not used by the queue. Job not submitted.", it means the hosts available to glastdataq have changed. The solution is to roll back the affected chunks (the doChunk streams) with a new value for HOSTLIST. When you roll back the streams from the frontend, on the confirmation page you are presented an opportunity to set or redefine variables. To figure out what the new value needs to be, do a "bqueues -l glastdataq". The output will include a line like "HOSTS: bbfarm/". In this case you'd enter HOSTLIST=bbfarm in the box on the confirmation page. bbfarm is actually a temporary thing for the cooling outage, when things get switched back to normal, the relevant line from bqueues will probably look more like "HOSTS: glastyilis+3 glastcobs+2 preemptfarm+1". Then the thing to enter in the box would be HOSTLIST="glastyilis glastcobs genfarm".

setCrashed skipped for successful stream

If everything goes well, then the setCrashed substream will be skipped. Makes sense to think about it, but could be confusing at first glance.