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.
The restorer can issue multiple calls to WriteAt in parallel. This can
result in unexpected orderings of the Truncate and WriteAt calls and
sometimes too short restored files.
We can either preallocate storage for a file or sparsify it. This
detects a pack file as sparse if it contains an all zero block or
consists of only one block. As the file sparsification is just an
approximation, hide it behind a `--sparse` parameter.
This writes files by using (*os.File).Truncate, which resolves to the
truncate system call on Unix.
Compared to the naive loop,
for _, b := range p {
if b != 0 {
return false
}
}
the optimized allZero is about 10× faster:
name old time/op new time/op delta
AllZero-8 1.09ms ± 1% 0.09ms ± 1% -92.10% (p=0.000 n=10+10)
name old speed new speed delta
AllZero-8 3.84GB/s ± 1% 48.59GB/s ± 1% +1166.51% (p=0.000 n=10+10)
`restic unlock` now only shows `successfully removed locks` if there were locks to be removed.
In addition, it also reports the number of the removed lock files.
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)
This is especially useful if ssh asks for a password or if closing the
initial connection could return an error due to a problematic server
implementation.
bazil/fuse expects us to return context.Canceled to signal that a
syscall was successfully interrupted. Returning a wrapped version of
that error however causes the fuse library to signal an EIO (input/output
error). Thus unwrap context.Canceled errors before returning them.
rclone can exit early for example when the connection to rclone is
relayed for example via ssh: `-o rclone.program='ssh user@example.org
forced-command'`
For backends which are able to atomically replace files, we just can
overwrite the old copy, if it is necessary to retry an upload. This has
the benefit of issuing one operation less and might be beneficial if a
backend storage, due to bugs or similar, could mix up the order of the
upload and delete calls.
When hard deleting the latest file version on B2, this uncovers earlier
versions. If an upload required retries, multiple version might exist
for a file. Thus to reliably delete a file, we have to remove all
versions of it.
Ignored packs were reported as an empty pack by EachByPack. The most
immediate effect of this is that the progress bar for rebuilding the
index reports processing more packs than actually exist.
Previously the buffer was grown incrementally inside `repo.LoadUnpacked`.
But we can do better as we already know how large the index will be.
Allocate a bit more memory to increase the chance that the buffer can be
reused in the future.
Instead of first checking whether a file is in the repository cache and
then opening it, we just can open the file. This saves one stat call. If
the file is in the cache, everything is fine and otherwise the code
follows its normal fallback path.
sort.Sort is not guaranteed to be stable. Go 1.19 has changed the
sorting algorithm which resulted in changes of the sort order. When
comparing snapshots with identical timestamp but different paths and
tags lists, there is not meaningful order among them. So just keep their
order stable.
Cleanly separate the directory presentation and the snapshot directory
structure. SnapshotsDir now translates the dirStruct into a format
usable by the fuse library and contains only minimal special case rules.
All decisions have moved into SnapshotsDirStructure which now creates a
fully preassembled tree data structure.
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.
Also make the errors a bit less verbose by not prepending the operation,
since pkg/xattr already does that. Old errors looked like
Listxattr: xattr.list /myfiles/.zfs/snapshot: invalid argument
pkg/sftp.Client.MkdirAll(d) does a Stat to determine if d exists and is
a directory, then a recursive call to create the parent, so the calls
for data/?? each take three round trips. Doing a Mkdir first should
eliminate two round trips for 255/256 data directories as well as all
but one of the top-level directories.
Also, we can do all of the calls concurrently. This may reintroduce some
of the Stat calls when multiple goroutines try to create the same
parent, but at the default number of connections, that should not be
much of a problem.
FutureBlob now uses a Take() method as a more memory-efficient way to
retrieve the futures result. In addition, futures are now collected
while saving the file. As only a limited number of blobs can be queued
for uploading, for a large file nearly all FutureBlobs already have
their result ready, such that the FutureBlob object just consumes
memory.
There is no real difference between the FutureTree and FutureFile
structs. However, differentiating both increases the size of the
FutureNode struct.
The FutureNode struct is now only 16 bytes large on 64bit platforms.
That way is has a very low overhead if the corresponding file/directory
was not processed yet.
There is a special case for nodes that were reused from the parent
snapshot, as a go channel seems to have 96 bytes overhead which would
result in a memory usage regression.
After the `BlobSaver` job is submitted, the buffer can be released and
reused by another `FileSaver` even before `BlobSaver.Save` returns. That
FileSaver will the change `buf.Data` leading to wrong backup statistics.
Found by `go test -race ./...`:
WARNING: DATA RACE
Write at 0x00c0000784a0 by goroutine 41:
github.com/restic/restic/internal/archiver.(*FileSaver).saveFile()
/home/michael/Projekte/restic/restic/internal/archiver/file_saver.go:176 +0x789
github.com/restic/restic/internal/archiver.(*FileSaver).worker()
/home/michael/Projekte/restic/restic/internal/archiver/file_saver.go:242 +0x2af
github.com/restic/restic/internal/archiver.NewFileSaver.func2()
/home/michael/Projekte/restic/restic/internal/archiver/file_saver.go:88 +0x5d
golang.org/x/sync/errgroup.(*Group).Go.func1()
/home/michael/go/pkg/mod/golang.org/x/sync@v0.0.0-20210220032951-036812b2e83c/errgroup/errgroup.go:57 +0x91
Previous read at 0x00c0000784a0 by goroutine 29:
github.com/restic/restic/internal/archiver.(*BlobSaver).Save()
/home/michael/Projekte/restic/restic/internal/archiver/blob_saver.go:57 +0x1dd
github.com/restic/restic/internal/archiver.(*BlobSaver).Save-fm()
<autogenerated>:1 +0xac
github.com/restic/restic/internal/archiver.(*FileSaver).saveFile()
/home/michael/Projekte/restic/restic/internal/archiver/file_saver.go:191 +0x855
github.com/restic/restic/internal/archiver.(*FileSaver).worker()
/home/michael/Projekte/restic/restic/internal/archiver/file_saver.go:242 +0x2af
github.com/restic/restic/internal/archiver.NewFileSaver.func2()
/home/michael/Projekte/restic/restic/internal/archiver/file_saver.go:88 +0x5d
golang.org/x/sync/errgroup.(*Group).Go.func1()
/home/michael/go/pkg/mod/golang.org/x/sync@v0.0.0-20210220032951-036812b2e83c/errgroup/errgroup.go:57 +0x91
Use runtime.GOMAXPROCS(0) as worker count for CPU-bound tasks,
repo.Connections() for IO-bound task and a combination if a task can be
both. Streaming packs is treated as IO-bound as adding more worker
cannot provide a speedup.
Typical IO-bound tasks are download / uploading / deleting files.
Decoding / Encoding / Verifying are usually CPU-bound. Several tasks are
a combination of both, e.g. for combined download and decode functions.
In the latter case add both limits together. As the backends have their
own concurrency limits restic still won't download more than
repo.Connections() files in parallel, but the additional workers can
decode already downloaded data in parallel.
Use only a single not completed pack file to keep the number of open and
active pack files low. The main change here is to defer hashing the pack
file to the upload step. This prevents the pack assembly step to become
a bottleneck as the only task is now to write data to the temporary pack
file.
The tests are cleaned up to no longer reimplement packer manager
functions.
Now with the asynchronous uploaders there's no more benefit from using
more blob savers than we have CPUs. Thus use just one blob saver for
each CPU we are allowed to use.
Previously, SaveAndEncrypt would assemble blobs into packs and either
return immediately if the pack is not yet full or upload the pack file
otherwise. The upload will block the current goroutine until it
finishes.
Now, the upload is done using separate goroutines. This requires changes
to the error handling. As uploads are no longer tied to a SaveAndEncrypt
call, failed uploads are signaled using an errgroup.
To count the uploaded amount of data, the pack header overhead is no
longer returned by `packer.Finalize` but rather by
`packer.HeaderOverhead`. This helper method is necessary to continue
returning the pack header overhead directly to the responsible call to
`repository.SaveBlob`. Without the method this would not be possible,
as packs are finalized asynchronously.
The short ids are not always unique. In addition, recovering from
damages is easier when having the full ids as that makes it easier to
access the corresponding files.
As MergeFinalIndex and index uploads can occur concurrently, it is
necessary for MergeFinalIndex to check whether the IDs for an index were
already set before merging it. Otherwise, we'd loose the ID of an index
which is set _after_ uploading it.
The GlobalOptions struct now embeds a backend.TransportOptions, so it
doesn't need to construct one in open and create. The upload and
download limits are similarly now a struct in internal/limiter that is
embedded in GlobalOptions.
There were three loops over the index in restic prune, to find
duplicates, to determine sizes (in pack.Size) and to generate packInfos.
These three are now one loop. This way, prune doesn't need to construct
a set of duplicate blobs, pack.Size doesn't need to contain special
logic for prune's use case (the onlyHdr argument) and pack.Size doesn't
need to construct a map only to have it immediately transformed into a
different map.
Some quick testing on a 160GiB local repo doesn't show running time or
memory use of restic prune --dry-run changing significantly.
... called backend/sema. I resisted the temptation to call the main
type sema.Phore. Also, semaphores are now passed by value to skip a
level of indirection when using them.
github.com/pkg/errors is no longer getting updates, because Go 1.13
went with the more flexible errors.{As,Is} function. Use those instead:
errors from pkg/errors already support the Unwrap interface used by 1.13
error handling. Also:
* check for io.EOF with a straight ==. That value should not be wrapped,
and the chunker (whose error is checked in the cases changed) does not
wrap it.
* Give custom Error methods pointer receivers, so there's no ambiguity
when type-switching since the value type will no longer implement error.
* Make restic.ErrAlreadyLocked private, and rename it to
alreadyLockedError to match the stdlib convention that error type
names end in Error.
* Same with rest.ErrIsNotExist => rest.notExistError.
* Make s3.Backend.IsAccessDenied a private function.
When given a buf that is big enough for a compressed blob but not its
decompressed contents, the copy at the end of LoadBlob would skip the
last part of the contents.
Fixes #3783.
This isn't doing anything. Channels should get cleaned up by the GC when
the last reference to them disappears, just like all other data
structures. Also inlined BufferPool.Put in Buffer.Release, its only
caller.
fd05037e1a changed the allocation batch
size from 256 to 128 under the assumption that an indexEntry is 60 bytes
on amd64, but it's 64: structs are padded out to a multiple of 8 for
alignment reasons. That means we'd waste no space in malloc even without
the batch allocation, at least on 64-bit machines. While that strategy
cuts the overallocation down dramatically for many small indexes, it also
seems to slow allocation down (Go 1.18, Linux, amd64, -benchtime=2s):
name old time/op new time/op delta
DecodeIndex-8 4.67s ± 5% 4.60s ± 1% ~ (p=0.953 n=10+5)
DecodeIndexParallel-8 4.67s ± 3% 4.60s ± 1% ~ (p=0.953 n=10+5)
IndexHasUnknown-8 37.8ns ± 8% 36.5ns ±14% ~ (p=0.841 n=5+5)
IndexHasKnown-8 38.5ns ±12% 37.7ns ±10% ~ (p=0.968 n=5+5)
IndexAlloc-8 615ms ±18% 607ms ± 1% ~ (p=1.000 n=10+5)
IndexAllocParallel-8 245ms ±11% 285ms ± 6% +16.40% (p=0.001 n=10+5)
MasterIndexAlloc-8 286ms ± 9% 275ms ± 2% ~ (p=1.000 n=10+5)
LoadIndex/v1-8 27.0ms ± 4% 26.8ms ± 1% ~ (p=0.690 n=5+5)
LoadIndex/v2-8 22.4ms ± 1% 22.8ms ± 2% +1.48% (p=0.016 n=5+5)
name old alloc/op new alloc/op delta
IndexAlloc-8 446MB ± 0% 446MB ± 0% -0.00% (p=0.000 n=8+4)
IndexAllocParallel-8 446MB ± 0% 446MB ± 0% -0.00% (p=0.008 n=8+5)
MasterIndexAlloc-8 213MB ± 0% 159MB ± 0% -25.47% (p=0.000 n=10+5)
name old allocs/op new allocs/op delta
IndexAlloc-8 913k ± 0% 2632k ± 0% +188.19% (p=0.008 n=5+5)
IndexAllocParallel-8 913k ± 0% 2632k ± 0% +188.21% (p=0.008 n=5+5)
MasterIndexAlloc-8 318k ± 0% 1172k ± 0% +267.86% (p=0.008 n=5+5)
Instead, this patch sets a batch size of 4, which means no space is
wasted by malloc on 64-bit and very little on 32-bit. It still gets very
close to the savings from not allocating in batches, without requiring
special code for bits.UintSize==64. Benchmark results, again for
Linux/amd64:
name old time/op new time/op delta
DecodeIndex-8 4.67s ± 5% 4.83s ± 9% ~ (p=0.315 n=10+10)
DecodeIndexParallel-8 4.67s ± 3% 4.68s ± 4% ~ (p=0.315 n=10+10)
IndexHasUnknown-8 37.8ns ± 8% 44.5ns ±19% ~ (p=0.095 n=5+5)
IndexHasKnown-8 38.5ns ±12% 36.9ns ± 8% ~ (p=0.690 n=5+5)
IndexAlloc-8 615ms ±18% 628ms ±18% ~ (p=0.218 n=10+10)
IndexAllocParallel-8 245ms ±11% 262ms ± 9% +7.02% (p=0.043 n=10+10)
MasterIndexAlloc-8 286ms ± 9% 287ms ±13% ~ (p=1.000 n=10+10)
LoadIndex/v1-8 27.0ms ± 4% 26.8ms ± 0% ~ (p=1.000 n=5+5)
LoadIndex/v2-8 22.4ms ± 1% 22.5ms ± 0% ~ (p=0.056 n=5+5)
name old alloc/op new alloc/op delta
IndexAlloc-8 446MB ± 0% 446MB ± 0% ~ (p=1.000 n=8+10)
IndexAllocParallel-8 446MB ± 0% 446MB ± 0% -0.00% (p=0.000 n=8+8)
MasterIndexAlloc-8 213MB ± 0% 160MB ± 0% -25.02% (p=0.000 n=10+9)
name old allocs/op new allocs/op delta
IndexAlloc-8 913k ± 0% 1333k ± 0% +45.94% (p=0.000 n=8+10)
IndexAllocParallel-8 913k ± 0% 1333k ± 0% +45.94% (p=0.000 n=8+8)
MasterIndexAlloc-8 318k ± 0% 525k ± 0% +64.99% (p=0.000 n=10+10)
The allocation method indexmap.newEntry has also been rewritten in a
form that is a few instructions shorter.
Apparently SMB/CIFS on Linux/macOS returns somewhat random errnos when
trying to sync a windows share which does not support calling fsync for
a directory.
This functionality has gone unused since
4b3dc415ef changed hashing.Reader's only
client to use ioutil.ReadAll on a bufio.Reader wrapping the hashing
Reader.
Reverts bcb852a8d0.
This removes RunWorkers, which had become mere overhead by successive
refactors. It also ensures that each former user of that function
returns any context error that occurs, so failure to complete an
operation is always reported as an error.
Apparently it can take a moment between closing a tempfile marked as
DELETE_ON_CLOSE and it actually being deleted. During that time the file
is inaccessible. Thus just skip deleting the temp file on windows.
Tree packs are cached locally at clients and thus benefit a lot from
being compressed. Ensure this be having prune always repack pack files
containing uncompressed trees.
A compressed index is only about one third the size of an uncompressed
one. Thus increase the number of entries in an index to avoid cluttering
the repository with small indexes.
