Event time: 18:07:12
Verbal Alert Time: 18:08:02
Tom Evans at LLO contacted me via TeamSpeak to verify the alarm was recieved by me as well.
After a dazzingly confusing narrative of what to do with the INJ_TRANS Guardian node in aLogs and E-mails, I selected to manually take the node to INJECT_KILL. I shall return the state to INJECT_SUCCESS after the requisite 1hr stand-down time. After speaking to Keita he assured me that this is ok to do and it won't harm or change anything but It is no longer neccesary to do manually. I spoke with Tom at LLO and he told me that they are still doind the manual Injections pause as well. I infer that they don't utilize the same method for handling this type of event that we do? I hope it isn't just me that is confused about this and why it isn't being handled identically by the entire project.
19:10UTC I've returned the above mentioned node to INJECT_SUCCESS. Below is a screenshot of what the node looked like when the alert occurred and then what it looked like after I returned it to INJECT_SUCCESS (left to right respectfully)
Minimal things you need to know about GRB alert:
Most important: You don't have to do anything on your own if the guardian is working as it should.
How to tell if the guardian is working correctly:
Go to gracedb and find the "event time". If it's still within one hour of the "event time", the injection guardian should be in EXTTRIG_ALERT_ACTIVE and the requested state (usually INJECT_SUCCESS) should remain untouched.
If you are out of this one hour window, the guardian should be back to WAIT_FOR_NEXT_INJECT (assuming that the requested state is INJECT_SUCCESS).
If the requested state is INJECT_KILL, guardian should be in INJECT_KILL no matter what.
If none of the above is true when there's no outstanding injection, the guardian is not working as it should. You should manually transition to INJECT_KILL, wait for an hour, and request INJECT_SUCCESS. Notify the LLO operator, write alog, and write an email.
What if I've just heard GRB alert but gracedb "event time" is already more than an hour ago, do I need to do something?
No. Or, if you're curious:
Go to gracedb and find the "event time" as well as the alert time labeled as "submitted". The latency here could be more than an hour on rare occasions.
The control room audible alarm is triggered after the "submitted" time which is by definition later than the "event time".
What if you requested INJECT_KILL after GRB alert when the guardian is fine:
No harm whatsoever, though not necessary.
Remember to request INJECT_SUCCESS later. If you confirm that the guardian is working as it should, you can request INJECT_SUCCESS even before one hour window expires, otherwise wait for an hour.
In the case of GRB reported by Ed, I see that the guardian was working as it should. After the alert was detected, the guardian jumped from WAIT_FOR_NEXT_INJECT to EXTTRIG_ALERT_ACTIVE without messing with the requested state which was INJECT_SUCCESS (his first screen shot).