For the experimental GHA remote caching (new in 2....
# general
p
For the experimental GHA remote caching (new in 2.20): https://www.pantsbuild.org/2.20/docs/using-pants/remote-caching-and-execution/remote-caching Should I stop caching the lmdb store? For reference, this is one of the workflows I'm adjusting: https://github.com/StackStorm/st2/blob/master/.github/workflows/pants.yaml#L32-L52 I'm wondering if I should set
cache-lmdb-store = 'false'
?
b
Yes! it’s fine-grained “remote” caching of that content so a course-grained upload/download of the whole LMBD directory is no longer necessary. Btw, I assume you’re aware of the GH limitation that may make the caching fail often: https://github.com/pantsbuild/pants/issues/20133
👍 1
p
Yup. I see those failures in my test runs. stats shows a bunch of remote writes with a few failures. Then the next time it runs, there are a bunch of successful reads and some more writes. So over time they will all end up getting cached I suspect.
👍 1
I wonder if the latency from all the tiny requests might end up taking more time than saving/retrieving the whole lmdb store... I hope your request for a bulk cache API gets some attention.
👍 1