This prevents an `rclone rcd` server from prematurely going into the
'deactivating' state, which was causing systemd to kill it with a
SIGABRT after the stop timeout.
Fixes#7540
This implements --auth-proxy for serve s3. In addition it:
* add listbuckets tests with and without authProxy
* use auth proxy test framework
* servetest: implement workaround for #7454
* update github.com/rclone/gofakes3 to fix race condition
This also
- move in use options (Opt) from vfsflags to vfscommon
- change os.FileMode to vfscommon.FileMode in parameters
- rework vfscommon.FileMode and add tests
Hugo has been making this warning for a while
WARN found no layout file for "html" for kind "term": You should
create a template file which matches Hugo Layouts Lookup Rules for
this combination.
This turned out to be the addition of the `groups:` keyword to the
command frontmatter. Hugo is doing something with this keyword though
this isn't documented in the frontmatter documentation.
The fix was removing the `groups:` keyword from the frontmatter since
it was never used by hugo.
On backends which return a valid object for "" with NewObject then
touch was going wrong as it thought it was passed an object.
This should not happen normally but s3 can be configured with
--s3-no-head where it is happy to believe that all objects exist.
Before this change, serve s3 did not consistently save the correct modtime value
in memory after putting or copying an object, which could sometimes cause an
incorrect modtime to be returned. This change fixes the issue by ensuring that
both "mtime" and "X-Amz-Meta-Mtime" are updated in b.meta when we have fresh data.
The issue was discovered on the TestBisyncRemoteRemote/ext_paths test.
Mentioned the possibility of skipping the symlink for new versions of
git-annex. (Probably deserves a test once the new git-annex trickles
down to CI platforms.)
I stopped trying to explain each config parameter here. Rather, the doc
now shows the user how to ask git-annex to describe config parameters
with `--whatelse`.
For each layout mode, these tests start with a git-annex-remote-rclone
remote, migrate it to a git-annex-remote-rclone-builtin remote. They
verify that a file copied pre-migration is still present and that `git
annex testremote` passes.
Issue #7625
Now that e2e tests are running in parallel, undoing the chdir to the
temp dir was causing flaky failures on cleanup. We don't need it anyway
because the worrisome subcommands have their working directory
controlled by `runInRepo()`.
Issue #7625
I'm hopeful that running these in parallel will not impact CI runtime
very much, but that likely depends on the number of CPU cores and
whether the tmp filesystem is backed by memory vs a physical disk.
Issue #7625
TestEndToEndRepoLayoutCompat exercises git-annex-remote-rclone-builtin
and git-annex-remote-rclone on the same rclone remote to ensure they are
compatible. It repeats the same test for all known layout modes.
Issue #7625
This commit adds support for the same repo layouts supported by
git-annex-remote-rclone. This should enable git-annex users with remotes
of type "rclone" to switch to a "rclone-builtin" without needing to
retransfer content.
Issue #7625
The .lck file filename length needs to be less than 255 bytes (not symbols) on
linux, and it was still too long on this test, because of the
subdir=測試_Русский_{spc}_{spc}_ě_áñ
on remotes with long names, such as TestChunkerChunk3bNoRenameLocal:
The .lck file filename length needs to be less than 255 bytes (not symbols) on
linux, and it was still too long on this test, because of the
subdir=測試_Русский_{spc}_{spc}_ě_áñ
Before this change, the decoder looked only for `io.EOF`, and if any other error
was returned, it could cause an infinite loop. This change fixes the issue by
breaking for any non-nil error.
This enables compatibility with versions of git-annex currently
available on GitHub's "ubuntu-latest" image, aka Ubuntu 22.04 Jammy.
Currently, Jammy is shipping git-annex 8.20210223-2ubuntu2.
https://packages.ubuntu.com/jammy/git-annex
Issue #7625
This commit implements milestone 2.1 for the gitannex subcommand:
https://github.com/rclone/rclone/issues/7625#issuecomment-1951403856
This rewrite makes a few improvements over the old shell script:
(1) It no longer uses the system's rclone.conf. Now, it writes the
rclone.conf file in an ephemeral directory.
(2) It no longer makes any assumptions about the contents of /tmp.
However, it now assumes that an rclone built from the HEAD commit is on
the PATH. It makes a best-effort attempt to verify this assumption, but
I'm not sure it's bulletproof.
I'm hoping that writing this in Go will enable more cross-platform
support in the future, but for now we're still restricted to Unixy
systems due to reliance on the HOME environment variable.
Issue #7625
Before this change, calling SetModTime on owncloud and nextcloud would
inadvertently erase the object's stored hashes. This change fixes the issue,
which was discovered by the bisync integration tests.
- fix parsing of connection string remotes (comma in name)
- skip remotes that can't upload empty files
- Mkdir the test case subdir before cache.Get-ing it
(only storj seems to need this... bug?)
Several fixes for the bisync integration tests:
- use unique initdir and datadir for each subtest so concurrent tests don't interfere with each other
- remove dots from dir names for bucket backends
- ignore messages specific to cache backend
- skip fix-case tests on backends that can't fix-case
- don't expect "{hashtype} differ" messages on backends with no hash types
- print timestamps in UTC local
More fixes will still be needed, but this should hopefully fix a good portion of them.
This change officially adds bisync to the nightly integration tests for all
backends.
This will be part of giving us the confidence to take bisync out of beta.
A number of fixes have been added to account for features which can differ on
different backends -- for example, hash types / modtime support, empty
directories, unicode normalization, and unimportant differences in log output.
We will likely find that more of these are needed once we start running these
with the full set of remotes.
Additionally, bisync's extremely sensitive tests revealed a few bugs in other
backends that weren't previously covered by other tests. Fixes for those issues
have been submitted on the following separate PRs (and bisync test failures will
be expected until they are merged):
- #7670 memory: fix deadlock in operations.Purge
- #7688 memory: fix incorrect list entries when rooted at subdirectory
- #7690 memory: fix dst mutating src after server-side copy
- #7692 dropbox: fix chunked uploads when size <= chunkSize
Relatedly, workarounds have been put in place for the following backend
limitations that are unsolvable for the time being:
- #3262 drive is sometimes aware of trashed files/folders when it shouldn't be
- #6199 dropbox can't handle emojis and certain other characters
- #4590 onedrive API has longstanding bug for conflictBehavior=replace in
server-side copy/move
This change adds the --direct-io flag to the mount. This means the
page cache is completely bypassed for reads and writes. No read-ahead
takes place. Shared mmap is disabled.
This is useful to accurately read files which may change length
frequently on the source.
This commit adds a new subcommand named "gitannex", aka
"git-annex-remote-rclone-builtin" when invoked via a symlink.
This accomplishes milestone 1 from issue #7625: "minimal support for the
external special remote protocol".
Issue #7625
Before this change, directory modtimes (and metadata) were always synced from
src to dst, even if already in sync (i.e. their modtimes already matched.) This
potentially required excessive API calls, made logs noisy, and was potentially
problematic for backends that create "versions" or otherwise log activity
updates when modtime/metadata is updated.
After this change, a new DirsEqual function is added to check whether dirs are
equal based on a number of factors such as ModifyWindow and sync flags in use.
If the dirs are equal, the modtime/metadata update is skipped.
For backends that require setDirModTimeAfter, the "after" sync is performed only
for dirs that could have been changed by the sync (i.e. dirs containing files
that were created/updated.)
Note that dir metadata (other than modtime) is not currently considered by
DirsEqual, consistent with how object metadata is synced (only when objects are
unequal for reasons other than metadata).
To sync dir modtimes and metadata unconditionally (the previous behavior), use
--ignore-times.
Before this change, the VFS layer did not properly handle unicode normalization,
which caused problems particularly for users of macOS. While attempts were made
to handle it with various `-o modules=iconv` combinations, this was an imperfect
solution, as no one combination allowed both NFC and NFD content to
simultaneously be both visible and editable via Finder.
After this change, the VFS supports `--no-unicode-normalization` (default `false`)
via the existing `--vfs-case-insensitive` logic, which is extended to apply to both
case insensitivity and unicode normalization form.
This change also adds an additional flag, `--vfs-block-norm-dupes`, to address a
probably rare but potentially possible scenario where a directory contains
multiple duplicate filenames after applying case and unicode normalization
settings. In such a scenario, this flag (disabled by default) hides the
duplicates. This comes with a performance tradeoff, as rclone will have to scan
the entire directory for duplicates when listing a directory. For this reason,
it is recommended to leave this disabled if not needed. However, macOS users may
wish to consider using it, as otherwise, if a remote directory contains both NFC
and NFD versions of the same filename, an odd situation will occur: both
versions of the file will be visible in the mount, and both will appear to be
editable, however, editing either version will actually result in only the NFD
version getting edited under the hood. `--vfs-block-norm-dupes` prevents this
confusion by detecting this scenario, hiding the duplicates, and logging an
error, similar to how this is handled in `rclone sync`.
Before this change, NOTICE log messages during bisync dry runs were unclear as
to the direction of the skipped operation (Path1 to 2 vs. 2 to 1.) This change
adjusts the cmd/bisync/log.go indent function to be more expressive about
direction.
Directory mod times are synced by default if the backend is capable
and directory metadata is synced if the --metadata flag is provided
and the backend is capable.
This updates the bisync golden tests also which were affected by
--dry-run setting of directory modtimes.
Fixes#6685
A consequence of this is that fs.Directory returned by the local
backend will now have a correct size in (rather than -1). Some tests
depended on this and have been fixed by this commit too.
Before this change, nfsmount ignored the --volname flag. After this change, the --
volname flag is respected, making it possible to set a custom volume name.
macOS users should note that Finder will show the correct volume name in most
places, but a notable exception is the sidebar, which will show "localhost".
This seems to be a system limitation (at least without `sudo`), but see the
discussion at https://github.com/rclone/rclone/issues/7503#issuecomment-1933997678
for some possible workarounds.
Before this change, if a user unmounted externally (for example, via the Finder
UI), rclone would not be aware of this and wait forever to exit -- effectively
causing a deadlock that would require Ctrl+C to terminate.
After this change, when the handler detects an external unmount, it calls a
function which allows rclone to cleanly shutdown the VFS and exit.
Before this change, writing files to an `nfsmount` via Finder on macOS would
cause critical errors, rendering `nfsmount` effectively unusable on macOS. This
change fixes the issue so that writes via Finder should be possible.
The issue was primarily caused by the handler's HandleLimit being set to -1. -1 is
the correct default for a NullAuthHandler, but not for a CachingHandler, which
interprets -1 not as "no limit" but as "no cache".
This change sets a high default of 1000000, and gives the user control over it
with a new --nfs-cache-handle-limit flag (available in both `serve nfs` and
`nfsmount`. A minimum of 5 is enforced, as any lower than this will be
insufficient to support directory listing.
The flushCache() function has a bug that causes it to never actually
flush the cache. Specifically, it checks whether DirCacheFlush is nil,
but never calls it.
The tests are already passing without flushing the dir cache, so this
commit just deletes flushCache() and its call sites.
Fixesrclone/rclone#7623
This change moves the --retries and --retries-sleep flags/variables from cmd to
config (consistent with --low-level-retries), so that they can be more easily
referenced from subcommands.
Before this change, in the event of a retryable error, bisync would always retry
the maximum number of times allowed by the `--retries` flag, even if one of the
retries was successful. This change fixes the issue, so that bisync moves on
after the first successful retry.
Before this change it wasn't possible to see where transfers were
going from and to in core/stats and core/transferred.
When use in rclone mount in particular this made interpreting the
stats very hard.
* serve restic: return internal error if listing failed
If listing a remote failed, then rclone returned http status "not
found". This has become a problem since restic 0.16.0 which ignores "not
found"-errors while listing a directory.
Just return internal server error, if something unexpected happens while
listing a directory.
* serve restic: fix error handling if getting a file fails
If the call to `newObject` in `serveObject` fails, then rclone always
returned a "not found" error. This prevents restic from distinguishing
permanent "not found" errors from everything else.
Thus, only return "not found" if the object is not found and an internal
server error otherwise.
Before this change, the path1 version of a file always prevailed during
--resync, and many users requested options to automatically select the winner
based on characteristics such as newer, older, larger, and smaller. This change
adds support for such options.
Note that ideally this feature would have been implemented by allowing the
existing `--resync` flag to optionally accept string values such as `--resync
newer`. However, this would have been a breaking change, as the existing flag
is a `bool` and it does not seem to be possible to have a `string` flag that
accepts both `--resync newer` and `--resync` (with no argument.) (`NoOptDefVal`
does not work for this, as it would force an `=` like `--resync=newer`.) So
instead, the best compromise to avoid a breaking change was to add a new
`--resync-mode CHOICE` flag that implies `--resync`, while maintaining the
existing behavior of `--resync` (which implies `--resync-mode path1`. i.e. both
flags are now valid, and either can be used without the other.
--resync-mode CHOICE
In the event that a file differs on both sides during a `--resync`,
`--resync-mode` controls which version will overwrite the other. The supported
options are similar to `--conflict-resolve`. For all of the following options,
the version that is kept is referred to as the "winner", and the version that
is overwritten (deleted) is referred to as the "loser". The options are named
after the "winner":
- `path1` - (the default) - the version from Path1 is unconditionally
considered the winner (regardless of `modtime` and `size`, if any). This can be
useful if one side is more trusted or up-to-date than the other, at the time of
the `--resync`.
- `path2` - same as `path1`, except the path2 version is considered the winner.
- `newer` - the newer file (by `modtime`) is considered the winner, regardless
of which side it came from. This may result in having a mix of some winners
from Path1, and some winners from Path2. (The implementation is analagous to
running `rclone copy --update` in both directions.)
- `older` - same as `newer`, except the older file is considered the winner,
and the newer file is considered the loser.
- `larger` - the larger file (by `size`) is considered the winner (regardless
of `modtime`, if any). This can be a useful option for remotes without
`modtime` support, or with the kinds of files (such as logs) that tend to grow
but not shrink, over time.
- `smaller` - the smaller file (by `size`) is considered the winner (regardless
of `modtime`, if any).
For all of the above options, note the following:
- If either of the underlying remotes lacks support for the chosen method, it
will be ignored and will fall back to the default of `path1`. (For example, if
`--resync-mode newer` is set, but one of the paths uses a remote that doesn't
support `modtime`.)
- If a winner can't be determined because the chosen method's attribute is
missing or equal, it will be ignored, and bisync will instead try to determine
whether the files differ by looking at the other `--compare` methods in effect.
(For example, if `--resync-mode newer` is set, but the Path1 and Path2 modtimes
are identical, bisync will compare the sizes.) If bisync concludes that they
differ, preference is given to whichever is the "source" at that moment. (In
practice, this gives a slight advantage to Path2, as the 2to1 copy comes before
the 1to2 copy.) If the files _do not_ differ, nothing is copied (as both sides
are already correct).
- These options apply only to files that exist on both sides (with the same
name and relative path). Files that exist *only* on one side and not the other
are *always* copied to the other, during `--resync` (this is one of the main
differences between resync and non-resync runs.).
- `--conflict-resolve`, `--conflict-loser`, and `--conflict-suffix` do not
apply during `--resync`, and unlike these flags, nothing is renamed during
`--resync`. When a file differs on both sides during `--resync`, one version
always overwrites the other (much like in `rclone copy`.) (Consider using
`--backup-dir` to retain a backup of the losing version.)
- Unlike for `--conflict-resolve`, `--resync-mode none` is not a valid option
(or rather, it will be interpreted as "no resync", unless `--resync` has also
been specified, in which case it will be ignored.)
- Winners and losers are decided at the individual file-level only (there is
not currently an option to pick an entire winning directory atomically,
although the `path1` and `path2` options typically produce a similar result.)
- To maintain backward-compatibility, the `--resync` flag implies
`--resync-mode path1` unless a different `--resync-mode` is explicitly
specified. Similarly, all `--resync-mode` options (except `none`) imply
`--resync`, so it is not necessary to use both the `--resync` and
`--resync-mode` flags simultaneously -- either one is sufficient without the
other.
Before this change, when a file was new/changed on both paths (relative to the
prior sync), and the versions on each side were not identical, bisync would
keep both versions, renaming them with ..path1 and ..path2 suffixes,
respectively. Many users have requested more control over how bisync handles
such conflicts -- including an option to automatically select one version as
the "winner" and rename or delete the "loser". This change introduces support
for such options.
--conflict-resolve CHOICE
In bisync, a "conflict" is a file that is *new* or *changed* on *both sides*
(relative to the prior run) AND is *not currently identical* on both sides.
`--conflict-resolve` controls how bisync handles such a scenario. The currently
supported options are:
- `none` - (the default) - do not attempt to pick a winner, keep and rename
both files according to `--conflict-loser` and
`--conflict-suffix` settings. For example, with the default
settings, `file.txt` on Path1 is renamed `file.txt.conflict1` and `file.txt` on
Path2 is renamed `file.txt.conflict2`. Both are copied to the opposite path
during the run, so both sides end up with a copy of both files. (As `none` is
the default, it is not necessary to specify `--conflict-resolve none` -- you
can just omit the flag.)
- `newer` - the newer file (by `modtime`) is considered the winner and is
copied without renaming. The older file (the "loser") is handled according to
`--conflict-loser` and `--conflict-suffix` settings (either renamed or
deleted.) For example, if `file.txt` on Path1 is newer than `file.txt` on
Path2, the result on both sides (with other default settings) will be `file.txt`
(winner from Path1) and `file.txt.conflict1` (loser from Path2).
- `older` - same as `newer`, except the older file is considered the winner,
and the newer file is considered the loser.
- `larger` - the larger file (by `size`) is considered the winner (regardless
of `modtime`, if any).
- `smaller` - the smaller file (by `size`) is considered the winner (regardless
of `modtime`, if any).
- `path1` - the version from Path1 is unconditionally considered the winner
(regardless of `modtime` and `size`, if any). This can be useful if one side is
usually more trusted or up-to-date than the other.
- `path2` - same as `path1`, except the path2 version is considered the
winner.
For all of the above options, note the following:
- If either of the underlying remotes lacks support for the chosen method, it
will be ignored and fall back to `none`. (For example, if `--conflict-resolve
newer` is set, but one of the paths uses a remote that doesn't support
`modtime`.)
- If a winner can't be determined because the chosen method's attribute is
missing or equal, it will be ignored and fall back to `none`. (For example, if
`--conflict-resolve newer` is set, but the Path1 and Path2 modtimes are
identical, even if the sizes may differ.)
- If the file's content is currently identical on both sides, it is not
considered a "conflict", even if new or changed on both sides since the prior
sync. (For example, if you made a change on one side and then synced it to the
other side by other means.) Therefore, none of the conflict resolution flags
apply in this scenario.
- The conflict resolution flags do not apply during a `--resync`, as there is
no "prior run" to speak of (but see `--resync-mode` for similar
options.)
--conflict-loser CHOICE
`--conflict-loser` determines what happens to the "loser" of a sync conflict
(when `--conflict-resolve` determines a winner) or to both
files (when there is no winner.) The currently supported options are:
- `num` - (the default) - auto-number the conflicts by automatically appending
the next available number to the `--conflict-suffix`, in chronological order.
For example, with the default settings, the first conflict for `file.txt` will
be renamed `file.txt.conflict1`. If `file.txt.conflict1` already exists,
`file.txt.conflict2` will be used instead (etc., up to a maximum of
9223372036854775807 conflicts.)
- `pathname` - rename the conflicts according to which side they came from,
which was the default behavior prior to `v1.66`. For example, with
`--conflict-suffix path`, `file.txt` from Path1 will be renamed
`file.txt.path1`, and `file.txt` from Path2 will be renamed `file.txt.path2`.
If two non-identical suffixes are provided (ex. `--conflict-suffix
cloud,local`), the trailing digit is omitted. Importantly, note that with
`pathname`, there is no auto-numbering beyond `2`, so if `file.txt.path2`
somehow already exists, it will be overwritten. Using a dynamic date variable
in your `--conflict-suffix` (see below) is one possible way to avoid this. Note
also that conflicts-of-conflicts are possible, if the original conflict is not
manually resolved -- for example, if for some reason you edited
`file.txt.path1` on both sides, and those edits were different, the result
would be `file.txt.path1.path1` and `file.txt.path1.path2` (in addition to
`file.txt.path2`.)
- `delete` - keep the winner only and delete the loser, instead of renaming it.
If a winner cannot be determined (see `--conflict-resolve` for details on how
this could happen), `delete` is ignored and the default `num` is used instead
(i.e. both versions are kept and renamed, and neither is deleted.) `delete` is
inherently the most destructive option, so use it only with care.
For all of the above options, note that if a winner cannot be determined (see
`--conflict-resolve` for details on how this could happen), or if
`--conflict-resolve` is not in use, *both* files will be renamed.
--conflict-suffix STRING[,STRING]
`--conflict-suffix` controls the suffix that is appended when bisync renames a
`--conflict-loser` (default: `conflict`).
`--conflict-suffix` will accept either one string or two comma-separated
strings to assign different suffixes to Path1 vs. Path2. This may be helpful
later in identifying the source of the conflict. (For example,
`--conflict-suffix dropboxconflict,laptopconflict`)
With `--conflict-loser num`, a number is always appended to the suffix. With
`--conflict-loser pathname`, a number is appended only when one suffix is
specified (or when two identical suffixes are specified.) i.e. with
`--conflict-loser pathname`, all of the following would produce exactly the
same result:
```
--conflict-suffix path
--conflict-suffix path,path
--conflict-suffix path1,path2
```
Suffixes may be as short as 1 character. By default, the suffix is appended
after any other extensions (ex. `file.jpg.conflict1`), however, this can be
changed with the `--suffix-keep-extension` flag (i.e. to instead result in
`file.conflict1.jpg`).
`--conflict-suffix` supports several *dynamic date variables* when enclosed in
curly braces as globs. This can be helpful to track the date and/or time that
each conflict was handled by bisync. For example:
```
--conflict-suffix {DateOnly}-conflict
// result: myfile.txt.2006-01-02-conflict1
```
All of the formats described [here](https://pkg.go.dev/time#pkg-constants) and
[here](https://pkg.go.dev/time#example-Time.Format) are supported, but take
care to ensure that your chosen format does not use any characters that are
illegal on your remotes (for example, macOS does not allow colons in
filenames, and slashes are also best avoided as they are often interpreted as
directory separators.) To address this particular issue, an additional
`{MacFriendlyTime}` (or just `{mac}`) option is supported, which results in
`2006-01-02 0304PM`.
Note that `--conflict-suffix` is entirely separate from rclone's main `--sufix`
flag. This is intentional, as users may wish to use both flags simultaneously,
if also using `--backup-dir`.
Finally, note that the default in bisync prior to `v1.66` was to rename
conflicts with `..path1` and `..path2` (with two periods, and `path` instead of
`conflict`.) Bisync now defaults to a single dot instead of a double dot, but
additional dots can be added by including them in the specified suffix string.
For example, for behavior equivalent to the previous default, use:
```
[--conflict-resolve none] --conflict-loser pathname --conflict-suffix .path
```
Before this change, certain shell quoting / escaping errors (particularly on
Windows) were not detected by Bisync, possibly resulting in incorrect expansion
and confusing errors. In particular, Windows paths with a single trailing
backslash followed by a quote would be interpreted as an escaped quote --
resulting in the quote and subsequent flags being erroneously considered part
of the path.
After this change, Bisync specifically checks for a few of the most common
patterns, and if detected, exits with a more helpful error message before doing
any damage.
Before this change, Bisync did not check to make sure that Path1 and Path2 do
not overlap, nor did it check for overlaps with `--backup-dir`. While `sync`
does check for these things, it can sometimes be fooled because of the way
Bisync calls it with `--files-from` filters. Relying on sync could also leave a
run in a half-finished state if it were to error in one direction but not the
other (`--backup-dir` only checks for overlaps with the dest.)
After this change, Bisync does its own check up front, so we can quickly return
an error and exit before any changes are made.
Background: Bisync uses lock files as a safety feature to prevent
interference from other bisync runs while it is running. Bisync normally
removes these lock files at the end of a run, but if bisync is abruptly
interrupted, these files will be left behind. By default, they will lock out
all future runs, until the user has a chance to manually check things out and
remove the lock.
Before this change, lock files blocked future runs indefinitely, so a single
interrupted run would lock out all future runs forever (absent user
intervention), and there was no way to change this behavior.
After this change, a new --max-lock flag can be used to make lock files
automatically expire after a certain period of time, so that future runs are
not locked out forever, and auto-recovery is possible. --max-lock can be any
duration 2m or greater (or 0 to disable). If set, lock files older than this
will be considered "expired", and future runs will be allowed to disregard them
and proceed. (Note that the --max-lock duration must be set by the process that
left the lock file -- not the later one interpreting it.)
If set, bisync will also "renew" these lock files every
--max-lock_minus_one_minute throughout a run, for extra safety. (For example,
with --max-lock 5m, bisync would renew the lock file (for another 5 minutes)
every 4 minutes until the run has completed.) In other words, it should not be
possible for a lock file to pass its expiration time while the process that
created it is still running -- and you can therefore be reasonably sure that
any _expired_ lock file you may find was left there by an interrupted run, not
one that is still running and just taking awhile.
If --max-lock is 0 or not set, the default is that lock files will never
expire, and will block future runs (of these same two bisync paths)
indefinitely.
For maximum resilience from disruptions, consider setting a relatively short
duration like --max-lock 2m along with --resilient and --recover, and a
relatively frequent cron schedule. The result will be a very robust
"set-it-and-forget-it" bisync run that can automatically bounce back from
almost any interruption it might encounter, without requiring the user to get
involved and run a --resync.
Before this change, bisync had no mechanism to gracefully cancel a sync early
and exit in a clean state. Additionally, there was no way to recover on the
next run -- any interruption at all would cause bisync to require a --resync,
which made bisync more difficult to use as a scheduled background process.
This change introduces a "Graceful Shutdown" mode and --recover flag to
robustly recover from even un-graceful shutdowns.
If --recover is set, in the event of a sudden interruption or other un-graceful
shutdown, bisync will attempt to automatically recover on the next run, instead
of requiring --resync. Bisync is able to recover robustly by keeping one
"backup" listing at all times, representing the state of both paths after the
last known successful sync. Bisync can then compare the current state with this
snapshot to determine which changes it needs to retry. Changes that were synced
after this snapshot (during the run that was later interrupted) will appear to
bisync as if they are "new or changed on both sides", but in most cases this is
not a problem, as bisync will simply do its usual "equality check" and learn
that no action needs to be taken on these files, since they are already
identical on both sides.
In the rare event that a file is synced successfully during a run that later
aborts, and then that same file changes AGAIN before the next run, bisync will
think it is a sync conflict, and handle it accordingly. (From bisync's
perspective, the file has changed on both sides since the last trusted sync,
and the files on either side are not currently identical.) Therefore, --recover
carries with it a slightly increased chance of having conflicts -- though in
practice this is pretty rare, as the conditions required to cause it are quite
specific. This risk can be reduced by using bisync's "Graceful Shutdown" mode
(triggered by sending SIGINT or Ctrl+C), when you have the choice, instead of
forcing a sudden termination.
--recover and --resilient are similar, but distinct -- the main difference is
that --resilient is about _retrying_, while --recover is about _recovering_.
Most users will probably want both. --resilient allows retrying when bisync has
chosen to abort itself due to safety features such as failing --check-access or
detecting a filter change. --resilient does not cover external interruptions
such as a user shutting down their computer in the middle of a sync -- that is
what --recover is for.
"Graceful Shutdown" mode is activated by sending SIGINT or pressing Ctrl+C
during a run. Once triggered, bisync will use best efforts to exit cleanly
before the timer runs out. If bisync is in the middle of transferring files, it
will attempt to cleanly empty its queue by finishing what it has started but
not taking more. If it cannot do so within 30 seconds, it will cancel the
in-progress transfers at that point and then give itself a maximum of 60
seconds to wrap up, save its state for next time, and exit. With the -vP flags
you will see constant status updates and a final confirmation of whether or not
the graceful shutdown was successful.
At any point during the "Graceful Shutdown" sequence, a second SIGINT or Ctrl+C
will trigger an immediate, un-graceful exit, which will leave things in a
messier state. Usually a robust recovery will still be possible if using
--recover mode, otherwise you will need to do a --resync.
If you plan to use Graceful Shutdown mode, it is recommended to use --resilient
and --recover, and it is important to NOT use --inplace, otherwise you risk
leaving partially-written files on one side, which may be confused for real
files on the next run. Note also that in the event of an abrupt interruption, a
lock file will be left behind to block concurrent runs. You will need to delete
it before you can proceed with the next run (or wait for it to expire on its
own, if using --max-lock.)
Before this change, bisync could only detect changes based on modtime, and
would refuse to run if either path lacked modtime support. This made bisync
unavailable for many of rclone's backends. Additionally, bisync did not account
for the Fs's precision when comparing modtimes, meaning that they could only be
reliably compared within the same side -- not against the opposite side. Size
and checksum (even when available) were ignored completely for deltas.
After this change, bisync now fully supports comparing based on any combination
of size, modtime, and checksum, lifting the prior restriction on backends
without modtime support. The comparison logic considers the backend's
precision, hash types, and other features as appropriate.
The comparison features optionally use a new --compare flag (which takes any
combination of size,modtime,checksum) and even supports some combinations not
otherwise supported in `sync` (like comparing all three at the same time.) By
default (without the --compare flag), bisync inherits the same comparison
options as `sync` (that is: size and modtime by default, unless modified with
flags such as --checksum or --size-only.) If the --compare flag is set, it will
override these defaults.
If --compare includes checksum and both remotes support checksums but have no
hash types in common with each other, checksums will be considered only for
comparisons within the same side (to determine what has changed since the prior
sync), but not for comparisons against the opposite side. If one side supports
checksums and the other does not, checksums will only be considered on the side
that supports them. When comparing with checksum and/or size without modtime,
bisync cannot determine whether a file is newer or older -- only whether it is
changed or unchanged. (If it is changed on both sides, bisync still does the
standard equality-check to avoid declaring a sync conflict unless it absolutely
has to.)
Also included are some new flags to customize the checksum comparison behavior
on backends where hashes are slow or unavailable. --no-slow-hash and
--slow-hash-sync-only allow selectively ignoring checksums on backends such as
local where they are slow. --download-hash allows computing them by downloading
when (and only when) they're otherwise not available. Of course, this option
probably won't be practical with large files, but may be a good option for
syncing small-but-important files with maximum accuracy (for example, a source
code repo on a crypt remote.) An additional advantage over methods like
cryptcheck is that the original file is not required for comparison (for
example, --download-hash can be used to bisync two different crypt remotes with
different passwords.)
Additionally, all of the above are now considered during the final --check-sync
for much-improved accuracy (before this change, it only compared filenames!)
Many other details are explained in the included docs.
Before this change, bisync used the "canonical" Fs name in the filename for its
listing files, including any {hexstring} suffix. An unintended consequence of
this was that if a user added a backend-specific flag from the command line
(thus "overriding" the config), bisync would fail to find the listing files it
created during the prior run without this flag, due to the path now having a
{hexstring} suffix that wasn't there before (or vice versa, if the flag was
present when the session was established, and later removed.) This would
sometimes cause bisync to fail with a critical error (if no listing existed
with the alternate name), or worse -- it would sometimes cause bisync to use an
old, incorrect listing (if old listings with the alternate name DID still
exist, from before the user changed their flags.)
After this change, the issue is fixed by always normalizing the SessionName to
the non-canonical version (no {hexstring} suffix), regardless of the flags. To
avoid a breaking change, we first check if a suffixed listing exists. If so, we
rename it (and overwrite the non-suffixed version, if any.) If not, we carry on
with the non-suffixed version. (We should only find a suffixed version if
created prior to this commit.)
The result for the user is that the same pair of paths will always use the same
.lst filenames, with or without backend-specific flags.
Before this change, integration tests often could not be run on backends with
differing features from the local system that goldenized them. In particular,
differences in modtime precision, checksum support, and encoding would cause
false positives. After this change, the tests more accurately account for the
features of the backend being tested, which allows us to see true positives
more clearly, and more meaningfully assess whether a backend is supported.
