Repository navigation
Add the serving cache store to the execution log (SpawnExec.origin)
#9943
Closed
ghostbuster91
started this conversation in
Ideas
Replies: 1 comment
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Problem
When setting up or debugging a remote cache, the question you need answered per task is which store served this hit. The execution log (
-Dsbt.experimental_execution_log=<path>) is the only machine-readable record of cache activity, but itsSpawnExecrecords say whether a task hit the cache, not where the hit came from:So
cacheHit: trueis ambiguous. A build that never successfully reaches the remote cache and merely hits its local disk store is indistinguishable in the log from one that reused a CI-populated remote entry. Those are the two outcomes you are trying to tell apart.Why the existing channel isn't sufficient
The per-command cache summary does distinguish stores:
but it has three limitations for this purpose:
NetworkClient.completeExecgates onbatchMode.get || !attached.get, so the server-sideAggregation.printSuccessoutput (which carriescacheSummary) is not what you see. CI generally runs--batch, so CI runs get no summary at all.The exec log is the channel that is independent of invocation mode, and it is one field short of answering the question.
Wdyt?
All reactions