S3 cloud container AttributeError when retrying an operation: S3Container instance has no attribute '_reactor'

[daira]

The retry handling for S3 containers in the cloud backend is broken.

In 2116#comment:27, we have a traceback ending with:

>   File \"/home/customer/LAFS_source/src/allmydata/storage/backends/cloud/cloud_common.py\", line 370, in _retry
>     d2 = self._handle_error(f, 1, None, description, operation, *args, **kwargs)
>   File \"/home/customer/LAFS_source/src/allmydata/storage/backends/cloud/cloud_common.py\", line 417, in _handle_error
>     d = task.deferLater(self._reactor, BACKOFF_SECONDS_BEFORE_RETRY[trynum-1], operation, *args, **kwargs)
> exceptions.AttributeError: S3Container instance has no attribute '_reactor'

This is a bug introduced during the refactoring that added allmydata.storage.backends.cloud.cloud_common.CommonContainerMixin. self._reactor is now set in the constructor inherited from CommonContainerMixin, but S3Container does not inherit from CommonContainerMixin and so does not call that constructor.

[daira]

Note that this is a classic example of the kind of error that would be caught by static typing but not necessarily by testing. (It does show that our test coverage for the S3 container is poor, which we knew about.)

[daira]

Pull request at ​https://github.com/LeastAuthority/tahoe-lafs/pull/4.

[daira]

The fix was in ​https://github.com/tahoe-lafs/tahoe-lafs/commit/de8b40188b67714a2a000206402a7458d6e55c68 and ​https://github.com/tahoe-lafs/tahoe-lafs/commit/e0ffe8bfcdb3f15f8e83d6472e058fa681bbe2a6 on the 1819-cloud-merge branch.

[warner]

Milestone renamed

[warner]

renaming milestone