clean up superevent on SIGNIFICANT_LABEL
Port over part of !1051 (merged) to fix #537 (closed)
The reason for #537 (closed) is as follows:
- EM_Selected freezes superevent. But it does not launch less-significant alerts yet. That will be done in !1051 (merged).
- In this patch, we port over the part of !1051 (merged) which cleans up the superevent on EM_SelectedConfident before sending the significant alert.
- Without this change, superevents that start out with a less significant G-event will have the alert not be sent due to failure to satisfy publishability condition: https://git.ligo.org/emfollow/gwcelery/-/blob/main/gwcelery/tasks/orchestrator.py#L700-707
- Another subtle but important change is needed in !1051 (merged) but refactored here: the
superevents.keyfunc
should explicitly order significant events higher than less-significant events i.e. a FAR=1e-5, SNR=12 is lower ranked than FAR=1e-10, SNR=11.
Merge request reports
Activity
assigned to @deep.chatterjee
requested review from @geoffrey.mo
added 1 commit
- 2e095829 - rank publishability condition of significant events above less-significant ones
I made a slight modification to first2years uploads for this particular superevent to recreate the case that is failing on production: https://gracedb-test.ligo.org/superevents/MS230223e/view/
Let me know if this makes sense @geoffrey.mo
added 1 commit
- a68f1bf9 - clean up superevent on SIGNIFICANT_LABEL, rank significant event higher
I'm looking at this but I want to note that your URL linked above (https://git.ligo.org/emfollow/gwcelery/-/blob/master/gwcelery/tasks/orchestrator.py#L577-588) links to the deprecated
master
branch, not the currentmain
branch. I think I found what you were referring to in the livemain
branch, but we should consider deleting or otherwise hiding themaster
branch to avoid this sort of confusion in the future.- Resolved by Geoffrey Mo
added 1 commit
- 76f151e1 - clean up superevent on SIGNIFICANT_LABEL, rank significant event higher
Thank you @geoffrey.mo. Made the changes requested, added comments, and also updated our sphinx docs. Please have a look.
mentioned in merge request !1051 (merged)