Closed mlincett closed 11 months ago
looks like you're good. if we have the space, let's use it
unless, you plan on adding a lot of files
At 14M, you could commit it directly to the repo. You start getting a warning at 50M, and github blocks anything larger than 100M.
We definitely could add this to prod-exe too. Any specific path you'd like it to appear as?
At 14M, you could commit it directly to the repo. You start getting a warning at 50M, and github blocks anything larger than 100M.
We definitely could add this to prod-exe too. Any specific path you'd like it to appear as?
As long as it is a single file. But if then we find the need to have multiple ones, a repo potentially approaching O(100M) in test data is not something I would be very happy with.
What's your take on git lfs, @dsschult ?
I generally dislike git lfs, unless there's a good reason to use it. I treat it like git submodules, which can have poor interactions.
for simweights i put the test data on our fileserver and used curl to download it https://github.com/icecube/simweights/blob/e892e8c0fc926e2f7b904aeba709a7ceb8dfdf99/.github/workflows/tests.yml#L37
Since the files for this are already available in /data/ana
and are not created ad-hoc, I guess I can use @kjmeagher approach. I guess I just need a repo admin to set secrets.ICECUBE_PASSWORD
for me.
That's already an org secret you should be able to use it arlready
@mlincett since you'll be pulling from an external source, can you add a step to check the checksum before the actual tests?
@mlincett since you'll be pulling from an external source, can you add a step to check the checksum before the actual tests?
A working example is now present in #7
Unfortunately, I don't know if there is a way to avoid downloading the same file for each python version.
Unfortunately, I don't know if there is a way to avoid downloading the same file for each python version.
GitHub Actions has caching tools that you can use. You'd create a preliminary step to download and put the file in your cache
It doesn't seem hard. But will require some time to test.
Just doing the download again might be as fast as the cache, because it's uploading and downloading it internally. We do have plenty of networking for that machine.
Fixed in #7, no caching for now.
I am thinking of adding test data for the
i3_to_json
script, however as an example/data/ana/realtime/alert_catalog_v2/input_files/Level2pass2_IC86.2011_data_Run00118435_Subrun00000000_00000144_event58198553.i3.zst
is 14M in size.Should we add this to the repository or can we think of adding test data to the
prod-exe
fileserver? cc @ric-evans @dsschult @briedelI would rather rely on an external source since this may be not the only test file we need to run.