jmckenzie-dev commented on PR #2554: URL: https://github.com/apache/cassandra/pull/2554#issuecomment-1673603525
> oh, one more feedback! We have the ability to list the tests, but how do we know about the `cost` of the tests? Right now Jenkins hard codes how many "splits" to have, and in CircleCI I try to get people to use my script that leverages a `cost` function to know how many resources are needed; it would be good to encode such a detail into this definition. > > Here is the super complex `cost` function I use > > ``` > unit: 20 > jvm_dtest: 4 > jvm_dtest_upgrade: 2 > ``` > > so how this works is `num splits = num_tests / cost`, this has been better at adapting to changes on the different branches and doesn't require we keep tuning CI... I want to forget about fixing CI! So... the higher the integer value for cost the... what, less splits you'll have right? Except that intersects with the number of tests? What's the value compared between jobs supposed to communicate independently? i.e. a 20 on unit tests vs. 4 on python dtests means... unit tests are 5x cheaper to run? Or 5x smaller on average than a python dtest? Or do those numbers not actually mean anything independently of the count of tests on their respective jobs? It's non-obvious to me why we'd have a _higher_ cost value for a unit test than a python dtest at face value. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
