conda / rattler

Rust crates to work with the Conda ecosystem.
BSD 3-Clause "New" or "Revised" License
264 stars 55 forks source link

Git LFS usage #798

Closed jaimergp closed 2 weeks ago

jaimergp commented 3 months ago

This repository is using a lot of LFS storage above the org quota:

image

I see that #512 tried to deal with this, but the history still has some. Maybe someone still had a local copy and accidentally reuploaded the content with a push for something else.

baszalmstra commented 3 months ago

As far as I understand, old commits (before #512) still contain references to LFS. We would have to rewrite the entire history to completely get rid of that which would be quite problematic for existing checkouts.

jaimergp commented 3 months ago

Yea, the script would need to export SHA-like tags for at least the release tags but it's not perfect. And this repo has a few tags already 😬 Just flagging this for awareness in case it was an omission. Not sure if it'll become a problem in the future or not.

baszalmstra commented 3 months ago

Yeah, we are aware but simply didn't want to rewrite history. Should we close this issue for now and reopen when it becomes a problem?

jaimergp commented 3 months ago

One "problem" already is the influx of billing warnings in the notifications for org admins. When this moves to conda it might incur in actual charges too.

baszalmstra commented 3 months ago

@jaimergp Do you have a suggestion on how we could "fix" this without having to rewrite all history?

@jezdez Is this a serious problem we have to tackle before graduating?

jaimergp commented 2 months ago

FYI, upon transfer we got this notification

Git LFS has been disabled on the organization conda-incubator because you’ve exceeded your data plan by at least 150%. Please purchase additional data packs to cover your bandwidth and storage usage:

https://github.com/organizations/conda-incubator/billing/data/upgrade

Current usage as of 13 Aug 2024 03:25PM CEST:

Bandwidth: 54.71 GB / 1 GB (5471%) Storage: 0.0 GB / 1 GB (0%)

wolfv commented 2 weeks ago

Is this still a problem?

jaimergp commented 2 weeks ago

I don't have admin access in this org, so I can't tell. Someone would need to go and visit https://github.com/organizations/conda/billing/data/upgrade to tell us. If you didn't do anything about it, the LFS traffic is still there and someone is paying for it.

wolfv commented 2 weeks ago

We did something about it at some point, but I am not sure if we still pull historic LFS data.

wolfv commented 2 weeks ago

Going to close this as I didn't hear any further from @jezdez or others.

jezdez commented 2 weeks ago

We're using 1.21 of 1 GB of data storage and 4 of 1 GB of traffic. So this isn't resolved as such, just not breaking the conda org. I'll buy a data pack now to cover for this.

jezdez commented 2 weeks ago

Done, this is now 50 GB for traffic and storage for the conda org per month. Please let me know if you need more, preferable ahead of time via a ticket in conda/infra so we have a paper trail.

Also kind reminder that these resources are shared with all conda org projects.

wolfv commented 2 weeks ago

Maybe we should add "lfs: false" to the ci pipelines