Re: Git maintenance fails without meaningful error message if any remote is no longer available

[Phillip Wood <phillip.wood123@gmail.com>]

Hi Anselm

On 05/05/2026 11:05, Anselm Schüler wrote:
> [ This is a duplicate message because I forgot to hit Reply All the last 
> time ]
> 
> Hi Phillip,
> 
> I think there may be a misunderstanding. My current systemd timers do 
> use --keep-going. The issue is that on the individual repo, git- 
> maintenance won’t fetch other remotes if one remote fails. --keep-going 
> will ensure that the other repos get processed, but the repo with the 
> failing remote won’t fetch any remotes after the failing one.
> 
> At least, that’s what appears to be happening from the output of the 
> command.

Oh sorry I'd misunderstood what was happening. I think that error comes 
from "git fetch --all" which dying in connect.c:die_initial_contact(). 
If there is one remote that is causing problems you could set 
remote.<name>.skipFetchAll so that git maintenance does not try and 
fetch from it. I've Cc'd Stolee to see if he has any better ideas for 
fixing this.

Thanks

Phillip

> On 05/05/2026 11:59, Phillip Wood wrote:
>> Hi Anselm
>>
>> On 30/04/2026 00:13, Anselm Schüler wrote:
>>> I have a repo with multiple remotes, one of which no longer exists. 
>>> When git-maintenance runs on it, it fails during the prefetch stage 
>>> because that remote doesn’t exist anymore, and gives a mostly 
>>> unhelpful error message:
>>>
>>> $ git maintenance run --schedule=daily
>>> ERROR: Repository not found.
>>> fatal: Could not read from remote repository.
>>>
>>> Please make sure you have the correct access rights
>>> and the repository exists.
>>> error: failed to prefetch remotes
>>> error: task 'prefetch' failed
>>>
>>> I think that
>>> 1. git-maintenance should report which remote it’s encountering an 
>>> error on
>>> 2. git-maintenance should continue fetching other remotes even if one 
>>> fails
>>
>> Since c75662bfc9 (maintenance: running maintenance should not stop on 
>> errors, 2024-04-24) which is in git 2.45.3 the systemd timer files 
>> installed by "git maintenance start" use "git for-each-repo --keep- 
>> going --config=..." to avoid this problem. Unfortunately we don't have 
>> a way to automatically upgrade the timer files for users who ran "git 
>> maintenance start" before that. I think if you run
>>
>>     git maintenance stop
>>     git maintenance start
>>
>> It will delete the old timer files and install the new ones. If that 
>> does not work you'll need to manually edit the files and add "--keep- 
>> going" to "git for-each-repo".
>>
>> Thanks
>>
>> Phillip
>>
>>> Now, on my system, the systemd timers for git-maintenance use git- 
>>> for- each-repo. Not sure if that’s upstream behaviour or something 
>>> Nix/home- manager does. But if it is upstream behaviour, it would 
>>> also be great to report the repo the error comes from, since I 
>>> basically had to guess right now which repo was erroring. Luckily I 
>>> have only three repos under maintenance so that was fine.
>>>
>>> Let me know if you agree that this should be done. I would be open to 
>>> writing a patch (no promises though)
>>>
>>> Anselm
>>>
>>>
>>
>