Closed mpermana closed 3 years ago
@danielfrg Looks like this is only lint error? And I can't figure out where the lint coming from.
Ah yeah, this is annoying, i think its because python version or something.
Could you commend that step on the Github actions yml?
This is going to require updating s3fs to be 0.5.0 or newer (that's when the loop
attribute was added)
Couple of questions. You dont necessarily have to know the answer here, but I figured I'd ask before trying to replicate this.
If it's at least reproducible with Minio we can potentially guard against regressions in the future without needing manual QA to check
edit: If you rebase onto current master branch, that linting error is fixed (well, disabled)
thanks for the PR btw!
Can't figure out why tests are still failing.
@danielfrg any idea how i can pass the test? its still failing
We also see this issue from time to time.
About the questions from @ericdill, the issue seems that we are doing too many calls to AWS S3 API. This could be easily measured profiling the python calls when using s3fs, maybe some cache layer for the metadata could save us here.
Can you add in the regression tests to generate a report with cPython with the amount of calls done from s3fs ?
This is going to require updating s3fs to be 0.5.0 or newer (that's when the
loop
attribute was added)Couple of questions. You dont necessarily have to know the answer here, but I figured I'd ask before trying to replicate this.
- Is this reproducible with GCP? Minio?
If it's at least reproducible with Minio we can potentially guard against regressions in the future without needing manual QA to check
- Any thoughts on how to test for improved performance on s3 / minio / GCS?
edit: If you rebase onto current master branch, that linting error is fixed (well, disabled)
any chance this gets merged to master? I am on s3 and while i only have ~30 files, i am still waiting approx 7 seconds for TTFB on each directory change in the file manager.
I know this took forever but I finally merged this changes in https://github.com/danielfrg/s3contents/pull/121.
I will be making a release soon.
Thank you very much for your contribution.
Performance is very bad to the point unusable when an S3 folder has many entries, say 1000 files.
When getting a list of S3 objects, we already have most of the information needed to build a directory model. So we can directly build the model.