[
https://issues.apache.org/jira/browse/YUNIKORN-3338?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dale Richardson updated YUNIKORN-3338:
--------------------------------------
Description:
With some cross-cutting performance improvements, it can very useful to be able
to provide evidence that there has been no functional change to the underlying
scheduling decisions made by Yunikorn.
One way of doing this is to check in a canonical trace file of scheduling
decisions, plus a testing harness that can remove all sources of randomness
from the decision processes, and drives a series of set scenarios through to
scheduling decisions. The testing can simply then be unit tests executing the
scenario and comparing to the previous checked in trace files. Either the
change in behaviour is deliberate, in which case an updated trace can be
generated and checked in, or the trace file can help pinpoint an in-advertant
change in scheduling behaviour
was:
With some cross-cutting performance improvements, it can very useful to be able
to provide evidence that there has been no functional change to the underlying
scheduling decisions made by Yunikorn.
One way of doing this is to check in a canonical trace file of scheduling
activities, plus a testing harness that can remove all sources of randomness
from the decision processes, and drives a series of set scenarios through to
scheduling decisions. The testing can simply then be unit tests executing the
scenario and comparing to the previous checked in trace files. Either the
change in behaviour is deliberate, in which case an updated trace can be
generated and checked in, or the trace file can help pinpoint an in-advertant
change in scheduling behaviour
> Golden test trace generation and validation
> -------------------------------------------
>
> Key: YUNIKORN-3338
> URL: https://issues.apache.org/jira/browse/YUNIKORN-3338
> Project: Apache YuniKorn
> Issue Type: Test
> Components: core - common
> Reporter: Dale Richardson
> Priority: Major
>
> With some cross-cutting performance improvements, it can very useful to be
> able to provide evidence that there has been no functional change to the
> underlying scheduling decisions made by Yunikorn.
> One way of doing this is to check in a canonical trace file of scheduling
> decisions, plus a testing harness that can remove all sources of randomness
> from the decision processes, and drives a series of set scenarios through to
> scheduling decisions. The testing can simply then be unit tests executing
> the scenario and comparing to the previous checked in trace files. Either
> the change in behaviour is deliberate, in which case an updated trace can be
> generated and checked in, or the trace file can help pinpoint an in-advertant
> change in scheduling behaviour
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
