Inconsistent results for repo API queries

[jbreckmckye]

Select Topic Area

Bug

Body

Hello all,

I'm having trouble fetching a list of my organization's repositories with both the REST and GraphQL APIs. The issue is with both, so I wonder if the problem is further up the chain.

In a nutshell: certain repos simply aren't returned by the API calls, despite paginating through the entire set; some are only returned sometimes, inconsistently. This seems the case despite using different PATs.

Here's an example call using the REST API via SDK:

const iterator = octokit.paginate.iterator(octokit.rest.repos.listForOrg, {
      type: 'private',
      direction: 'asc',
      sort: 'full_name',
      org: ORGANIZATION_NAME,
      per_page: 100,
    })

This returns 733 responses for my org today - but in the UI, 833...

And not just that - several of the items returned by the iterator are incorrect. There's a whole bunch of what appear to be forks of our code by non-org users, which isn't what I understand private to mean here.

My first assumption was that they key I was using might not have enough scope - but even using my own issued PAT, the query is missing org repositories that I created and am an admin of.

The project I'm working on needs a reliable list of repositories, in order to mirror state and capture important metrics about code at our organisation. What can I do to get this consistent list of organisation repositories?

---

EDIT

One thing I discovered is that if I stopped using my existing PAT, and instead created a brand new personal access token with exactly the same scopes, now the query seems to retrieve a larger (possibly full?) set of results.

Is it possible there is something odd happening with caches?

[jbreckmckye]

Hi @alixconstantin, thanks for getting back so quickly.

I did consider some of these options, including the possibility of the token being problematic.

One thing I discovered is that if I stopped using my existing PAT, and instead created a brand new personal access token with exactly the same scopes, now the query seems to retrieve a larger (possibly full?) set of results.

Is it possible there is something odd happening with caches?

[Raicuparta]

Hey @jbreckmckye, did you ever come to any conclusion here? I think I might have a similar problem fetching releases.

Also, that other reply you got is ChatGPT so uh, yeah.

[DropSnorz]

Hello @Raicuparta . I don't know if it's related to your problem, but I noticed I have some "ghost" releases in one of my repositories.

https://github.com/orgs/community/discussions/76962

[jbreckmckye]

No, I ended up just renewing the token. For a while after if I used the old token repos would be missing, so I figured something must be being cached somewhere

[uladzimirbabrou-tomtom]

I'm using Search API (by private GitHub App) to list repos of our organisation and getting different results all the time.
Even if I use search box on the organisation home page to find repos, I get 177, 169, 166, 157, 154, 150, ... results for the same input string almost every time.

It makes impossible to rely on GitHub Search API even for Enterprise users.

[asocorrocognex]

Anyone noticing a similar kind of inconsistency with the CoPilot metrics API? We are getting different values for the same dates.