Fix Version/s: None
DM-29776 revealed that the Gen3 version of ForcedPhotCcdTask uses much more memory than ForcedPhotCoadd, which doesn't make sense given that (for HSC) detectors are half the size of patches and we don't need HeavyFootprints or replace-with-noise in ForcedPhotCcd.
- we're using deferred loading instead of grabbing all overlapping patch catalogs at once;
- we're not not loading HeavyFootprints unnecessarily;
- we're not doing anything else stupidly memory-hungry.
If this ticket fixes the problem, please make sure the new requestMemory values are communicated to operators and the rest of the team.
- relates to
DM-29776 Attempt complete two-QG HSC RC2 run on w16 on one tract
Arun Kannawadi added a comment -
Looks good. It might be safer to run ci_hsc also on Jenkins. Not sure if Jim/Nate said it was unnecessary.
Arun Kannawadi added a comment - Looks good. It might be safer to run ci_hsc also on Jenkins. Not sure if Jim/Nate said it was unnecessary.
This merge actually only addresses the first of Jim's three points in the ticket description; if this does not solve the issue, I'll file another follow-up ticket.
(Changes done at pair coding with Jim Bosch and Nate Lust)