{{ message }}
Repository navigation
feat: set Content-MD5 header to avoid silent data corruption - #815
Merged
bhearsum merged 1 commit intoAug 11, 2026
Merged
Conversation
In [bug 2060889](https://bugzilla.mozilla.org/show_bug.cgi?id=2060889) we're looking at some chain of trust verification failures caused by hash mismatches between `chain-of-trust.json` and the files in GCS. After a great deal of debugging, I ended up finding that [internal aiohttp retries on PUT requests can send incomplete data](aio-libs/aiohttp#13329). The short version of that bug is that the file handle is advanced every time a request is prepared, and if a failure+retry happens after that, it advances to the next chunk of data rather than resending the failed chunk. GCS uploads [support the Content-MD5 header](https://docs.cloud.google.com/storage/docs/data-validation#xml-single-request-upload), which will cause a 400 error to be returned if the uploaded data does not match the md5 provided in the header. Setting this will let us retry at our level if an upload is corrupt in this specific scenario, but more importantly, for any other future scenario that may come up. This means that we don't benefit from aiohttp's retries, which might save a bit of bandwidth and time, but it's a more general solution. A different way to work around this would be avoid passing plain file handles and either pass the data (impractical due to the size of the files), or one of aiohttp's Payload objects. The latter seems to work, although they're not well documented, and I don't understand it enough to feel confident in doing so.
bhearsum
marked this pull request as ready for review
August 5, 2026 18:15
ahal
approved these changes
Aug 7, 2026
ahal
left a comment
ahal
left a comment
Contributor
There was a problem hiding this comment.
Sounds reasonable!
Seems like a pretty bad bug in aiohttp though, which is surprising. Maybe we're not using an up-to-date version.
Contributor
Author
bhearsum
added a commit
that referenced
this pull request
Aug 12, 2026
This is causing permafailures: ``` public/logs/chain_of_trust.log was corrupted in transit: the storage backend rejected the body as not matching Content-MD5 +r5pkJPuO/JWdjABXz6gEQ==. Retrying. ``` I'm guessing the md5 calculation is mismatched somehow. It's probably not worth debugging considering aiohttp merged a fix upstream (aio-libs/aiohttp#13330), and the issue this was attempting to work around is so rare.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
In bug 2060889 we're looking at some chain of trust verification failures caused by hash mismatches between
chain-of-trust.jsonand the files in GCS. After a great deal of debugging, I ended up finding that internal aiohttp retries on PUT requests can send incomplete data. The short version of that bug is that the file handle is advanced every time a request is prepared, and if a failure+retry happens after that, it advances to the next chunk of data rather than resending the failed chunk.GCS uploads support the Content-MD5 header, which will cause a 400 error to be returned if the uploaded data does not match the md5 provided in the header. Setting this will let us retry at our level if an upload is corrupt in this specific scenario, but more importantly, for any other future scenario that may come up. This means that we don't benefit from aiohttp's retries, which might save a bit of bandwidth and time, but it's a more general solution.
A different way to work around this would be avoid passing plain file handles and either pass the data (impractical due to the size of the files), or one of aiohttp's Payload objects. The latter seems to work, although they're not well documented, and I don't understand it enough to feel confident in doing so.