1

Let's say I have pushed a branch containing a 100 MB binary blob file to my AWS CodeCommit by mistake, which has now made cloning the repo very slow.

I have deleted the branch containing it, but cloning the repo is still slow, and the binary blob is still there even though I have verified it's unreachable from any branch or commit.

How do I force AWS CodeCommit to garbage-collect it? Or at least, can I count on CodeCommit to eventually garbage-collect it (in two weeks, or whatever time interval it uses)?

sashoalm
  • 75,001
  • 122
  • 434
  • 781

2 Answers2

1

Or at least, can I count on CodeCommit to eventually garbage-collect it (in two weeks, or whatever time interval it uses)?

Yes, if the blob is truly unreachable, CodeCommit should garbage collect it eventually. Reading the docs for git gc:

When common porcelain operations that create objects are run, they will check whether the repository has grown substantially since the last maintenance, and if so run git gc automatically.

So assuming that CodeCommit executes the code that does this, say by directly executing git add from the command line, then it will run garbage collection automatically.

Code-Apprentice
  • 81,660
  • 23
  • 145
  • 268
  • "say by directly executing git add from the command line" - could you clarify that? I'm asking because I don't have SSH access to CodeCommit. I can execute `git add` on my local repo, but I can only push to the remote CodeCommit server. – sashoalm Jan 20 '21 at 08:14
  • @sashoalm I'm not familiar with CodeCommit, but I assume it has a web interface similar to those that I use with GitHub or BitBucket. When I wrote those particular words, I had in mind that when you take actions through that interface, CodeCommit executes shell commands on its server. For example, when you make a pull request and then click the "merge" button, under the hood it probably executes a `git merge` command. – Code-Apprentice Jan 20 '21 at 16:06
  • @sashoalm Even a `git push` will "create objects" in the remote repository. So the way I read the documentation, `git gc` will eventually run. – Code-Apprentice Jan 20 '21 at 16:07
0

I found the issue, it was with the script from Which commit has this blob?. It was listing commits that reference the blob, but ONLY IF those commits are in the currently checked out branch.

My blob was being referenced by another (different) branch! Once I removed that, the blob disappeared immediately from AWS CodeCommit, so I was wrong to think CodeCommit keeps blobs around even when not referenced.

sashoalm
  • 75,001
  • 122
  • 434
  • 781