Closed mswiatlo closed 3 years ago
For what it's worth I've adjusted my restart script to just do:
pandamon -d 31 -i failed -a | awk '{print $2}' | panda-resub-taskid -n 1
And this works fine. So the -a option is already retrieving the right information.
This probably makes sense. I blame @lawrenceleejr because the thing returned by -t
is apparently 'reqid'
, which sounds like another thing entirely. Larry was this your idea or mine? Why did we do it? Can we change it to 'jeditaskid'
?
Anyway, sounds like the Danny Antrim option saves the day again. I mean, why would you ever not use the @dantrim option?
This probably makes sense. I blame @lawrenceleejr because the thing returned by
-t
is apparently'reqid'
, which sounds like another thing entirely. Larry was this your idea or mine? Why did we do it? Can we change it to'jeditaskid'
?
Regardless we should make sure this is better documented as
$ pandamonium --version
pandamonium, version 0.2.1
$ pandamon --help | grep taskid
-t, --taskid output only taskid. useful for piping.
doesn't make this distinction clear.
Totally agree. Looks like my fault. Indeed the reqid used to work in practice but probably should have always been the actual task ID. Almost definitely me not understanding the difference (still).
Almost definitely me not understanding the difference (still).
You aren't alone @lawrenceleejr. Is there public documentation on the difference between the two?
How about I just change it to give the Jedi taskid?
Yeah, I don’t think the other one is useful for anything anymore.
(And this isn’t anyone here’s fault per se— both used to work, and the change was unannounced as usual. Normal panda development fun!)
Is there public documentation on the difference between the two?
From the PanDA JEDI docs
JEDITASKID
is the unique task identifier in JEDI. It comes fromATLAS_DEFT.PRODSYS2_TASK_ID_SEQ
independently ofT_TASK_REQUEST.REQID
. UntilDEFT
is ready task requests come fromAKTR
T_TASK_REQUEST.REQID
is stored inJEDI_TASKS.REQID
which is mapped tojob.TASKID
.
So not super clear, but at least draws a distinciton.
HI,
I recently found that restarting jobs via the TaskID returned by pbook (with -t) has started to fail; this number apparently instead is corresponding to the PandaID. PBook used to be smart and accept either the TaskID or PandaID for the restart of the job, but now it really is demanding the TaskID. Can we change what is returned? Asoka said:
Does this request make sense or am I doing something wrong?
Cheers, Max