Ambiguity between "flat" and "multinode-aware" job IDs.
As known, all LAVA jobs have "real", DB-assigned integer IDs. However, for sub-jobs of multinode jobs, a "virtual" ID of the format <lead_job_id>.<small_integer> is used here and there (underlying resolved to the real ID).
The problem is exactly in "here and there" part, multinode-aware IDs are used inconsistently. Particular usecase:
- The job https://validation.linaro.org/scheduler/job/2290833.1 failed due to serial output timeout.
- My immediate thought is "maybe this particular board is broken".
- So, I copy job's ID as shown, "2290833.1", to the clipboard, and click on device name to open page which shows history of jobs which ran on this device: https://validation.linaro.org/scheduler/device/lite-frdm-k64f-02
- I then press Ctrl+F and paste copied job ID into the browser search field, to see what happened around that job.
- And I find nothing, because that page happens to use real DB job IDs!
So, this is a request to:
- Update device info page, like https://validation.linaro.org/scheduler/device/lite-frdm-k64f-02 , to show multinode-aware ID's.
- But also request for LAVA maintainers to grep thru page templates which show job ID's, and see if some other pages should be updated too.
I assume that if multinode-aware IDs were introduced in the first place, and shown here and there, then it's the right direction of change to show them as consistently as possible.
If however there're some technical difficulties with that, e.g. resolving real IDs to multinode IDs would be too slow for long job lists, then it may make sense to go in the opposite direction - to show real IDs consistently. I don't want to propose to drop multinode-aware IDs completely, but on a page like https://validation.linaro.org/scheduler/job/2290833.1 , please show both real ID and multinode-aware ID (clearly labelled which is which), so operations like listed above can performed efficiently by users.
Thanks!