Display priority values as specified by user (removing the negative sign)

[SultanOrazbayev]

Only the first element in the tuple is negated (it contains the user-specified priority), other components are unchanged.

This is a follow-up from #4185

[jrbourbeau]

Thanks for the PR @SultanOrazbayev. Apologies for the CI build failures, they are unrelated to the changes in this PR. I've merged the main branch into this branch which should fix the test failures (hope that's okay)

[mrocklin]

Only the first element in the tuple is negated (it contains the user-specified priority), other components are unchanged.

Hrm, that's interesting. We probably can't negate only one element, otherwise the presented value doesn't actually represent the priority. We probably have to negate them all or none.

[SultanOrazbayev]

Uh oh. I thought just the first element should be negated because of these lines:

distributed/distributed/scheduler.py

Lines 1967 to 1970 in debe1b8

	for key in set(priority) & touched_keys:
	ts = self.tasks[key]
	if ts.priority is None:
	ts.priority = (-(user_priority.get(key, 0)), generation, priority[key])

In the dashboard right now (without this commit), the negative sign appears only on the user-specified values... so I thought this would fix it.

[mrocklin]

I suspect that the other priorities are just genuinely negative values. It doesn't matter so much to the system if we use negative or positive values. It matters if things are higher or lower than each other. By reversing just some of these values it throws things out of alignment

…

On Tue, Nov 3, 2020 at 7:38 AM SultanOrazbayev ***@***.***> wrote: Uh oh. I thought just the first element should be negated because of these lines: https://github.com/dask/distributed/blob/debe1b80e74e43ed88001c40a9cf07cfefa2fc47/distributed/scheduler.py#L1967-L1970 In the dashboard right now (without this commit), the negative sign appears only on the user-specified values... so I thought this would fix it. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4211 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACKZTG2RFPP4WNH756JOT3SOAPV5ANCNFSM4TGHESLA> .