Ergonomic, hermetic, user-friendly software build system for Python, Go, Java, Kotlin, Protobuf, Scala, Shell. Pants lets you fearlessly scale up your codebase!

Pants

but the fingerprint bit is where the actual invalidation happens.

I wouldn't be opposed to adding another axis to the existing invalidation sorcery

rather than putting it in the directory name?

sorry about the confusion guys I should have updated the fingerprint in the first place when I wrote this code.

nah, not your fault... we all missed it in review

Sorry ran out the door after I dropped that vague statement

I meant that right now cache keys are three parts: fingerprint, transitive fingerprint, global cache key

We could have fingerprint, transitive fingerprint, task cache key, global cache key

I'm setting up another jenkins for our inhouse CI pipeline at QuizUp -- the old one has been getting really slow over time, and I'm really interested in setting up a distributed cache for our pants builds. Would anyone be up for sharing stack/configs for me get going with that?

<@U0EH3DM7W>: our distributed cache uses NFS.  The configuration is pretty simple:

```
[cache]
# '/mnt/nfs/shared-cache/pants-artifact-cache' is available on ec2 workers
read_from: ["/mnt/nfs/shared-cache/pants-artifact-cache"]
read: True
write_to: ["/mnt/nfs/shared-cache/pants-artifact-cache"]
write: True
# Turn off auto-purging of the buildcache.  We have a cronjob that cleans the NFS mount
# for all artifacts older than 2 weeks
max_entries_per_target: 0
```

Here is the nginx configuration we used to use: <https://docs.google.com/document/d/1_RMg8yvURMmewG-dVwl5Z9cSJfejKokecgCD075J2iI/edit#heading=h.jzlpkido5b1k>

DistributedArtifactcachinginPantsatSquare.txt

<@U053H6LQS>: we have a kind of file system sharing for the artifacts currently, and I feel like we're having random build failures because of that

<@U053H6LQS>: what's the point of the nginx setup? For developers only, since the ec2 build workers use nfs directly?

<@U0EH3DM7W>:Before we had NFS, we used NGINX.   Currently, we don't have devs hooked up to the distributed cache, but that is the way we would do it.

<@U053H6LQS>: do you never see errors in build due to some kind of FS operation conflicts?

For NFS, no we don't.  The way the buildcache work is well aligned with NFS.  It writes data to a temp file, then moves it into place and never changes the file again.

I think Foursquare has a cache document (and possibly more) we would be willing to share. Let me chat with some folks and I will get back to you.