With most cloud providers, traffic is much more expensive than API
calls. Thus slightly bias streamPack towards a bit more API calls in
exchange for slightly less traffic.
Due to the interface of streamPack, we cannot guarantee that operations
progress fast enough that the underlying connections remains open. This
introduces partial failures which massively complicate the error
handling.
Switch to a simpler approach that retrieves the pack in chunks of 32MB.
If a blob is larger than this limit, then it is downloaded separately.
To avoid multiple copies in memory, an auxiliary interface
`discardReader` is introduced that allows directly accessing the
downloaded byte slices, while still supporting the streaming used by the
`check` command.
The current in-memory index becomes stale after prune or repair index
have run. Thus, just drop the in-memory index altogether once these
commands have finished.
Some tests have to explicitly create pack files with blobs that don't
match their ID. For those blobs the builtin verification of the
repository must be disabled.
LoadBlobsFromPack is now part of the repository struct. This ensures
that users of that method don't have to deal will internals of the
repository implementation.
The filerestorer tests now also contain far fewer pack file
implementation details.
If an error occurred while streaming a pack file, this could result in
passing some of the blobs multiple times to the callback function. This
significantly complicates using StreamPack correctly and is unnecessary.
Retries do not change the content of a blob and thus only deliver the
same result over and over again.
As the `Fatal` error type only includes a string, it becomes impossible
to inspect the contained error. This is for a example a problem for the
fuse implementation, which must be able to detect context.Canceled
errors.
Co-authored-by: greatroar <61184462+greatroar@users.noreply.github.com>
This method had a buffer argument, but that was nil at all call sites.
That's removed, and instead LoadUnpacked now reuses whatever it
allocates inside its retry loop.
The retry backend does not return the original error, if its execution
is interrupted by canceling the context. Thus, we have to manually
ensure that the invalid data error gets returned.
Additionally, use the retry backend for some of the repository tests, as
this is the configuration which will be used by restic.
The retry printed the filename twice:
```
Load(<lock/04804cba82>, 0, 0) returned error, retrying after 720.254544ms: load(<lock/04804cba82>): invalid data returned
```
now the warning has changed to
```
Load(<lock/04804cba82>, 0, 0) returned error, retrying after 720.254544ms: invalid data returned
```
The Test method was only used in exactly one place, namely when trying
to create a new repository it was used to check whether a config file
already exists.
Use a combination of Stat() and IsNotExist() instead.
Repositories with mixed packs are probably quite rare by now. When
loading data blobs from a mixed pack file, this will no longer trigger
caching that file. However, usually tree blobs are accessed first such
that this shouldn't make much of a difference.
The checker gets a simpler replacement.
Sparse files contain large regions containing only zero bytes. Checking
that a blob only contains zeros is possible with over 100GB/s for modern
x86 CPUs. Calculating sha256 hashes is only possible with 500MB/s (or
2GB/s using hardware acceleration). Thus we can speed up the hash
calculation for all zero blobs (which always have length
chunker.MinSize) by checking for zero bytes and then using the
precomputed hash.
The all zeros check is only performed for blobs with the minimal chunk
size, and thus should add no overhead most of the time. For chunks which
are not all zero but have the minimal chunks size, the overhead will be
below 2% based on the above performance numbers.
This allows reading sparse sections of files as fast as the kernel can
return data to us. On my system using BTRFS this resulted in about
4GB/s.
Sending data through a channel at very high frequency is extremely
inefficient. Thus use simple callbacks instead of channels.
> name old time/op new time/op delta
> MasterIndexEach-16 6.68s ±24% 0.96s ± 2% -85.64% (p=0.008 n=5+5)
For large pack sizes we might be only interested in the first and last
blob of a pack file. Thus stream a pack file in multiple parts if the
gaps between requested blobs grow too large.
