This change removes the hardcoded Google auth mechanism for the GCS
backend, instead using Google's provided client library to discover and
generate credential material.
Google recommend that client libraries use their common auth mechanism
in order to authorise requests against Google services. Doing so means
you automatically support various types of authentication, from the
standard GOOGLE_APPLICATION_CREDENTIALS environment variable to making
use of Google's metadata API if running within Google Container Engine.
In #1590, it was mentioned that while lines read from exclude files via
`--exclude-file` have leading and trailing spaces stripped, this is not
the case for lines read via `--files-from`. This commit fixes that,
spaces are always stripped.
This commit removes the bandwidth displayed during backup process. It is
misleading and seldomly correct, because it's neither the "read
bandwidth" (only for the very first backup) nor the "upload bandwidth".
Many users are confused about (and rightly so), c.f. #1581, #1033, #1591
We'll eventually replace this display with something more relevant when
#1494 is done.
During the development of #1524 I discovered that the Google Cloud
Storage backend did not yet use the HTTP transport, so things such as
bandwidth limiting did not work. This commit does the necessary magic to
make the GS library use our HTTP transport.
When looking up a blob in the master index, with several
indexes present in the master index, a significant amount of time
is spent generating errors for each failed lookup. However, these
errors are often used to check if a blob is present, but the contents
are not inspected making the overhead of the error not useful.
Instead, change Index.Lookup (and Index.LookupSize) to instead return
a boolean denoting if the blob was found instead of an error. Also change
all the calls to these functions to handle the new function signature.
benchmark old ns/op new ns/op delta
BenchmarkMasterIndexLookupSingleIndex-6 820 897 +9.39%
BenchmarkMasterIndexLookupMultipleIndex-6 12821 2001 -84.39%
BenchmarkMasterIndexLookupSingleIndexUnknown-6 5378 492 -90.85%
BenchmarkMasterIndexLookupMultipleIndexUnknown-6 17026 1649 -90.31%
benchmark old allocs new allocs delta
BenchmarkMasterIndexLookupSingleIndex-6 9 9 +0.00%
BenchmarkMasterIndexLookupMultipleIndex-6 59 19 -67.80%
BenchmarkMasterIndexLookupSingleIndexUnknown-6 22 6 -72.73%
BenchmarkMasterIndexLookupMultipleIndexUnknown-6 72 16 -77.78%
benchmark old bytes new bytes delta
BenchmarkMasterIndexLookupSingleIndex-6 160 160 +0.00%
BenchmarkMasterIndexLookupMultipleIndex-6 3200 240 -92.50%
BenchmarkMasterIndexLookupSingleIndexUnknown-6 1232 48 -96.10%
BenchmarkMasterIndexLookupMultipleIndexUnknown-6 4272 128 -97.00%
Is seems that #1307 is similar to #1087, which describes a comparable
observation on Apple's new filesystem. #1389 Has been committed and
fixes the problem on Darwin.
Although I'm not sure if the root cause of the issue is the same the
solution is similar for OpenBSD, and leverages #1389.
On Darwin, allow a 1μs difference in restored timestamps, because
macOS <10.13 cannot restore with nanosecond precision and the current
version of Go (1.9.2) does not yet support the new syscall required
for this. (#1087#1389)
Add --last flag to snapshots command to only show the last entry for any
(hostname, paths) combination.
This makes it easier to check when various paths were last backed up.
Backup was choosing a parent snapshot that had the same tags, which
makes backup unnecessarily slow when there are newer snapshots with
different tags.
There's no reason parent has to have the same tags.
This change makes backup choose the newest snapshot instead.
This removes the conditions that checks if the password is supplied
through environment variable or file and outputs password is successful
on terminal and when --quiet is not supplied.
This adds some feedback when entering the password on the command line.
When the password is entered and supplied through stdin (and stdout is a
terminal) then the a message saying `password is correct` if correct is
printed.
This adds additional output to the check command when no errors were
found. It means that when all checks have been completed, the following
output is displayed:
No errors were found
The output is added to make sure that it is easier to understand that no
errors were found.
Full example output:
Create exclusive lock for repository
Load indexes
Check all packs
Check snapshots, trees and blobs
No errors were found
This commit removes the `manpages` and `autocomplet` commands and
replaces them with the more generic `generate` command. Also, zsh
completion file support was added.
This commits adds rudimentary support for a cache directory, enabled by
default. The cache directory is created if it does not exist. The cache
is used if there's anything in it, newly created snapshot and index
files are written to the cache automatically.
This was a bit tricky: We start the ssh binary, but we want it to ignore
SIGINT. In contrast, restic itself should process SIGINT and clean up
properly. Before, we used `setsid()` to give the ssh process its own
process group, but that means it cannot prompt the user for a password
because the tty is gone.
So, now we're passing in two functions that ignore SIGINT just before
the ssh process is started and re-install it after start.
The option is named --exclude-if-present and accepts a parameter
filename[:content]. Directories are excluded and their contents is not
backed up if they contain a file with the specified name and,
optionally, that starts with the specified content. The tagfile itself
is never excluded.
There is also a shortcut --exclude-caches that works in the same way as
the likewise-named option of tar(1): Directories are recognized as cache
if they contain a file named "CACHEDIR.TAG.
Closes #317.
By default (i.e., without profile.NoShutdownHook), profile.Start listens
for SIGINT and will stop the profile and call os.Exit(0).
restic already listens for SIGINT and runs its own cleanup handlers
before calling os.Exit(0).
As is, these handlers are racing when an interrupt occurs, and in my
experience, restic tends to win the race, resulting in an unusable
profile.
Eliminate the race and properly stop profiles on interrupt by disabling
package profile's signal handler and instead stop the profile in a
restic cleanup handler.
An exclude filter is basically a 'wildcard but foo', so even if a
childMayMatch, other children of a dir may not, therefore childMayMatch
does not matter, but we should not go down unless the dir is selected
for restore.
This improves restore performance by several orders of magniture by not
going through the whole tree recursively when we can anticipate that no
match will ever occur.
Instead of determining the password lazily during ReadPassword(), do so now in
cobra.PersistentPreRunE() so we can store the result in the globalOptions and
reuse/override when applicable without having to worry about the environment
or flag options interfering.
As `restic key add` uses the same `ReadPasswordTwice()` as the
rest of restic, it is sensitive to the environment variable
RESTIC_PASSWORD or --password-file= override.
When asking for the new key, temporary remove these 2 overrides, forcing
the password to be asked.
Switch from a function passed as a parameter to a cleanup function,
which is also executed when the test function panics, so no temporary
directories are left behind.
Errors returned from backend.LoadAll() were not handled, leading to
these fatal errors from the unpacker trying to read the size from the end of
an empty buffer:
`seeking to read header length failed: bytes.Reader.Seek: negative position`
This change takes care of returning on error, as well as showing which pack
failed to load and validating pack integrity.
Before, the fuse integration test was run and the tests were never
finished, because the testing code did not detect any errors when the
fusermount binary returned an error. This commit fixes it.
It was observed that a restic repository still contained overlapping
indexes after `rebuild-index` has been called. This is caused by
instantly forgetting that blobs have already been saved once a full
index has been written during index rebuilding.
This commit adds a (failing) test that shows the behaviour.
The method of determining if a repository exists doesn't work on Windows, since
the "url.Scheme" will contain the drive letter - "c" in "c:\backup",
so as a first step we check if the URL can be opened as a file,
and if so, we assume it is a 'local' type repository.
This changes `repository.LoadBlob()` so that a destination buffer must
be provided, which enables the fuse code to use a buffer from a
`sync.Pool`. In addition, release the buffers when the file is closed.
At the moment, the max memory usage is defined by the max file size that
is read in one go (e.g. with `cat`). It could be further optimized by
implementing a LRU caching scheme.
Since backend.ID is always a slice of constant length, use an array
instead of a slice. Mostly, arrays behave as slices, except that an
array cannot be nil, so use `*backend.ID` insteaf of `backend.ID` in
places where the absence of an ID is possible (e.g. for the Subtree of a
Node, which may not present when the node is a file node).
This change allows to directly use backend.ID as the the key for a map,
so that arbitrary data structures (e.g. a Set implemented as a
map[backend.ID]struct{}) can easily be formed.
This adds the exclude patterns used to create a backup in the snapshot,
so we can later compute statistics (like git does) on the data
structure, e.g. added/removed files etc. For that, we need the exclude
pattern.
