This patch provides the support of synchronous cache space recovery
to allow read threads to recover from ENOSPC errors when cache space
can be recovered from cache items that are not in use or safe to be
reset/emptied .
The patch complements the existing cache cleaning process in two ways.
Firstly, the existing cache cleaning process is time-driven that runs
periodically. The cache space can run out while the cache cleaner
thread is still waiting for its next scheduled run. The io threads
encountering ENOSPC return an internal error to the applications
in this case even when cache space can be recovered to avoid this
error. This patch addresses this problem by having the read threads
kick the cache cleaner thread in this condition to recover cache
space preventing unnecessary ENOSPC errors from being seen by the
applications.
Secondly, this patch enhances the cache cleaner to support cache
item reset. Currently the cache purge process removes cache
items that are not in use. This may not be sufficient when the
total size of the working set exceeds the cache directory's
capacity. Like in the current code, this patch starts the purge
process by removing cache files that are not in use. Cache items
whose access times are older than vfs-cache-max-age are removed first.
After that, other not-in-use items are removed in LRU order until
vfs-cache-max-size is reached. If the vfs-cache-max-size (the quota)
is still not reached at this time, this patch adds a cache reset
step to reset/empty cache files that are still in use but not
dirtied. This enables application processes to continue without
seeing an error even when the working set depletes the cache space
as long as there is not a large write working set hoarding the
entire cache space.
By design this patch does not add ENOSPC error recovery for write
IOs. Rclone does not empty a write cache item until the file data
is written back to the backend upon close. Allowing more cache
space to be consumed by dirty cache items when the cache space is
already running low would increase the risk of exhausting the cache
space in a way that the vfs mount becomes unreadable.
The deadlock was caused in transfermap.go by calling mu.RLock() in one
function then calling it again in a sub function. Normally this is
fine, however this leaves a window where mu.Lock() can be called. When
mu.Lock() is called it doesn't allow the second mu.RLock() and
deadlocks.
Thead 1 Thread 2
String():mu.RLock()
del():mu.Lock()
sortedSlice():mu.RLock() - DEADLOCK
Lesson learnt: don't try using locks recursively ever!
This patch fixes the problem by removing the second mu.RLock(). This
was done by factoring the code that was calling it into the
transfermap.go file so all the locking can be seen at once which was
ultimately the cause of the problem - the code which used the locks
was too far away from the rest of the code using the lock.
This problem was introduced in:
bfa5715017 fs/accounting: sort transfers by start time
Which hasn't been released in a stable version yet
- add a directory to the optional Purge interface
- fix up all the backends
- add an additional integration test to test for the feature
- use the new feature in operations.Purge
Many of the backends had been prepared in advance for this so the
change was trivial for them.
This is preparation for getting the Accounting to check the context,
buf first we need to get it in place. Since this is one of those
changes that makes lots of noise, this is in a seperate commit.
go1.15 introduced a stricter policy for what you can convert with
`string()` and now `go vet` warns if you try to do `string(int)`.
See: https://github.com/golang/go/issues/32479
If the parameter being passed is an object then it can be passed as a
JSON string rather than using the `--json` flag which simplifies the
command line.
rclone rc operations/list fs=/tmp remote=test opt='{"showHash": true}'
Rather than
rclone rc operations/list --json '{"fs": "/tmp", "remote": "test", "opt": {"showHash": true}}'
Before this change, if the cache was given a source `remote:file` it
stored `remote:` with the error `fs.ErrorIsFile` attached. This meant
that if it `remote:` was subsequently looked up it would return the
`fs.ErrorIsFile` error.
This broke `moveto remote:file remote:file2` as moveto would lookup
`remote:` from the second argument and erroneously get the
`fs.ErrorIsFile` error.
This likely broke other commands too.
This was broken in
4c9836035 fs/cache: Add Pin and Unpin and canonicalised lookup
Which was released in v1.52.0
The fix is to make a new cache entry for `remote:` with no error
attached in the case that the original call returned `fs.ErrorIsFile`.
This adds the missing WSAECONNREFUSED error to the list of errors we
can retry under Windows.
> Connection refused. No connection could be made because the target
> computer actively refused it.
It also adds any relevant errors I could see in the error code list.
See: https://forum.rclone.org/t/failing-to-upload-large-file-to-b2/17085
This was caused by using the stats group from the context passed in by the rcd
rather than the global stats group.
Signed-off-by: Gary Kim <gary@garykim.dev>
Before this change `--track-renames-strategy` was broken. The hashing
method it used could declare times that were very close together to be
different.
The time hash was discarded and instead we check the modification time
window on every hash match.
Provided that the user doesn't use `--track-renames-strategy` on a
huge number of identically sized files this will perform just fine.
See: https://forum.rclone.org/t/track-renames-strategy-modtime-doesnt-work/16992/5
This adds expire and unlink fields to the PublicLink interface.
This fixes up the affected backends and removes unlink parameters
where they are present.
Before this change you could use "" as a valid remote, so `rclone lsf
""` would work. This was treated as the current directory.
This is unexpected and creates a footgun for scripting when an empty
variable is passed to rclone by accident.
This fix returns the error "can't use empty string as a path" instead
of allowing it.
Before this change there was some ambiguity about whether passwords
were obscured on not passing them into config create or config update.
This change adds the --obscure and --no-obscure flags to make the
intent clear.
It also updates the remote control and the tests.
Fixes#3728
If your filenames contain two near-identical Unicode characters,
rclone will normalize these, making them identical. This flag
gives you the ability to keep them unique. This might
create unintended side effects, such as duplicating files that
contain certain Unicode characters, when downloading them from
certain cloud providers to a macOS filesystem.
Fixes#4228
During a copy/sync command, if an operation fails due to a network
issue and is retried, the underlying io.Reader is re-initialised,
but the stats for bytes already read are not reset, leading to incorrect
stats. THis was fixed by resetting the bytes read when an Account is
re-initialized.
This includes a new directory listing template which was originally
from the Caddy project (used with permission and copyright attribution).
This is used whenever we serve directory listings so `rclone serve
http`, `rclone serve webdav` and `rclone rcd --rc-serve`
This also modifies the tests so they work with the original template which
is easier to debug.
Bind rclone standard input to password command's standard input. This
allows to provide password from a pipe and collect it using cat.
The typical use case is when rclone is on a remote server with an
encrypted configuration. This solved the environment variable
issue (#3368) and the password storage on remote host.
Now the following chain is allowed:
echo 'secret' | ssh host.example.com \
sudo -u rclone \
rclone --config /path/to/rclone.conf \
--password-command 'cat' ls remote:
Signed-off-by: Sébastien Gross <seb•ɑƬ•chezwam•ɖɵʈ•org>
Co-authored-by: Sébastien Gross <seb•ɑƬ•chezwam•ɖɵʈ•org>
if running `rclone rcd --rc-user=admin --rc-pass=admin
--rc-allow-origin="*"`, lots of duplicate warnings apperent in log
Warning: Allow origin set to *. This can cause serious security problems.
Warning: Allow origin set to *. This can cause serious security problems.
....
This is not conducive to analyzing debugging info.
Therefore, let's show it only once.
Before this change we stored cached Fs under the config string the
user gave us. As the result of fs.ConfigString() can often be
different after the backend has canonicalised the paths this meant
that we could not look up backends in the cache reliably.
After this change we store cached Fs under their config string as
returned from fs.ConfigString(f) after the Fs has been created. We
also store a map of user to canonical names (where they are different)
so the users can look up Fs under the names they passed to rclone too.
This change along with Pin and Unpin is necessary so we can look up
the Fs in use reliably in the `backend/command` remote control
interface.
Before this change if you specified --hash MD5 in rclone lsf it would
calculate all the hashes and just return the MD5 hash which was very
slow on the local backend.
Likewise specifying --hash on rclone lsjson was equally slow.
This change introduces the --hash-type flag (and corresponding
internal parameter) so that the hashes required can be selected in
lsjson.
This is used internally in lsf when the --hash parameter is selected
to speed up the hashing by only hashing with the one hash specified.
Fixes#4181
These commands are for implementing backend specific
functionality. They have documentation which is placed automatically
into the backend doc.
There is a simple test for the feature in the backend tests.
Before this fix, FixRangeOption would substitute RangeOptions it
wanted to get rid of with with empty HTTPOption. This caused a problem
now that backends interpret HTTPOptions.
This fix subsitutes those with NullOptions which aren't interpreted as
HTTPOptions. This patch also improves the unit tests.
This allows rclone to exit with a non-zero return code if no files are
transferred. This is useful when calling rclone as part of a workflow/script
pipeline as it allows the end user to stop processing if no files have been
transferred.
NB: Enabling this option will return in rclone exiting non-zero if there are no
transfers. Depending on how your're currently using rclone in your scripts,
this may break existing setups!
--files-from parses input files by ignoring comments starting with # and ;
and stripping whitespace from start and end of strings.
The --files-from-raw flag was added that reads every line from the file ignoring
comment characters and not stripping whitespace while maintaining
backwards compatibility.
Fixes#3762
Before this change if there were two files with the same name and the
same ID in the same directory, dedupe would delete one of them but
since these are actually the same file (with the same ID) then both
files would be deleted leading to data loss.
This should never actually happen, however it did happen as part of a
bug introduced in rclone which was fixed by
dfc7215bf9 drive: fix duplicate items when using --drive-shared-with-me #4018
This change checks to see if any of the duplicates have the same ID
and if they do it refuses to delete them.
Before this change these tests attempted to measure transfers and
checks in lieu of having a rename statistic with a very complicated
heuristic.
The change switches over to using the rename statistic which should be
100% reliable.
Before this change the first pass of --delete-before would output
"There was nothing to transfer" and then proceed to transfer things.
This makes sure the message isn't printed in the delete phase.
See: https://forum.rclone.org/t/incorrect-debug-output/15267
This commit corrects the logic for --track-renames-strategy which
broke the integration tests.
It also improves the parsing of the argument and adds a test for that.
This commit adds the `--track-renames-strategy` flag which allows the
user to choose the strategy for tracking renames when using the
`--track-renames` flag.
This can be "hash" or "modtime" or both currently.
This, when used with `--track-renames-strategy modtime` enables
support for tracking renames in encrypted remotes.
Fixes#3696Fixes#2721
Before this change backends which introduce overhead (eg crypt) were
failing to upload the first file.
This change increases the threshold to 2k to allow the first file to
go through even with some overhead but the next file to definitely
fail.
Before this change we checked the transfer was out of range only
before the Read call. This means that we returned all the data to the
reader before declaring an error. This means that some backends wrote
the file even though an error was returned.
This fix checks the transfer after the Read as well, and chops the
excess characters off the read data if we are over the limit so that
we don't ever deliver all the data.
This fixes the tests introduced as part of 6f1766dd9e and #2672
on backends other than local.
Before this change the exit code for transfer limit exceeded was
incorrect. This was because the `resolveExitCode` function unwraps the
error thus reading the underlying error which is not the same as the
error it was comparing to (`ErrorMaxTransferLimitReached`).
This change fixes it by splitting the error definition in two so that
when the Fatal error is unwrapped we match against
`ErrorMaxTransferLimitReached` however when we return the error we
return `ErrorMaxTransferLimitReachedFatal`.
In bde0334bd8 "operations: fix setting the timestamp on Windows
for multithread copy" the test for multithread copy failed to take
into account the modify window of the remote under test.
Before this fix we attempted to set the modification time on the file
when it was open. This works fine on Linux but not on Windows. The
test was also incorrect testing the source file rather than the
destination file.
This closes the file before setting the modification time and fixes
the tests.
Fixes#3994
Before this change, the elapsed time shown with the --progress flag
would not print ".0s" so the elapsed time.
This change will make it so that the line width is kept a bit more
consistent by always printing to a fixed-point.
This does change the displayed value when the elapsed time
is less than 1s, in which it used to be that the value would be shown
in ms or smaller units.
Signed-off-by: Gary Kim <gary@garykim.dev>
Before this changed we unconditionally fetched the MimeType. On Some
backends like s3 and swift this takes an extra transaction which meant
that `lsf` on those backends was needlessly slow.
This adds an internal option so `lsf` can declare whether it wants
MimeTypes or not depending on whether the user asked for them and an
external flag `--no-mimetype` for `lsjson`.
See: https://forum.rclone.org/t/reliably-setup-incremental-updates/14006/8
This is a timing dependent test and to make it long enough so that it
would work with the remotes would make it too long for local tests.
The code paths are identical for local vs non-local so just run on
local.
This fixes the integration tests.
This gives you more control over how long rclone will run for, making
it easier to script backups, e.g. via cron. Once the `--max-duration`
time limit is reached, no new transfers will be initiated, but those
already in-flight will be allowed to complete.
Fixes#985
Before this change the expect/continue timeout was set to
--conntimeout which was 60s by default which is too long to wait.
This was noticed when using s3 with a proxy which apparently didn't
support expect / continue properly.
Set --expect-continue-timeout 0 to disable expect/continue.
Statistics of ransfers which were interrupted are not cleared before
retry iteration. These transfers completed with over 100 percentage.
This change clears transfer accounting before next retry iteration is
done in order to keep numbers in track.
Fixes#3861
Without the fix we can have a race, example:
```
Write at 0x00c000432039 by goroutine 187:
github.com/rclone/rclone/fs/accounting.(*StatsInfo).Error()
fs/accounting/stats.go:495 +0x3f1
github.com/rclone/rclone/fs/accounting.(*StatsInfo).Error-fm()
fs/accounting/stats.go:477 +0x55
github.com/rclone/rclone/fs/walk.listRwalk.func1()
fs/walk/walk.go:162 +0xd2
github.com/rclone/rclone/fs/walk.walk.func2()
fs/walk/walk.go:402 +0x30f
Previous read at 0x00c000432039 by goroutine 184:
github.com/rclone/rclone/fs/accounting.(*statsGroups).sum()
fs/accounting/stats_groups.go:351 +0xcae
github.com/rclone/rclone/fs/accounting.rcTransferredStats()
fs/accounting/stats_groups.go:132 +0x1f4
```
Fixes#3844
Before this change if an operation was retried on operations.Copy and
the operation was large enough to use an async buffer then an async
buffer was leaked on the retry. This leaked memory, a file handle and
a go routine.
After this change if Account.WithBuffer is called and there is already
a buffer, then a new one won't be allocated.
rclone library users might be intrested in changing default value to
other, or even disabling it. With current version it's impossible which
leads to races when number of uploaded objects exceeds default limit.
Fixes#3732
For few commands, RClone counts a error multiple times. This was fixed by
creating a new error type which keeps a flag to remember if the error has
already been counted or not. The CountError function now wraps the original
error eith the above new error type and returns it.
from rsync manual:
--compare-dest=DIR
This option instructs rsync to use DIR on the destination machine as an
additional hierarchy to compare destination files against doing transfers
(if the files are missing in the destination directory). If a file is found
in DIR that is identical to the sender's file, the file will NOT be
transferred to the destination directory. This is useful for creating
a sparse backup of just files that have changed from an earlier backup.
--copy-dest=DIR
This option behaves like --compare-dest, but rsync will also copy unchanged
files found in DIR to the destination directory using a local copy.
This is useful for doing transfers to a new destination while leaving
existing files intact, and then doing a flash-cutover when all files
have been successfully transferred.
On google fs (drive, google photos, and google cloud storage), if
headless is selected, do not open browser.
This also supplies a new option "auth-no-open-browser" for authorize
if the user does not want it.
This should fix#3323.
This was broken in e337cae0c5 when we deleted the transfers
immediately.
This is fixed by keeping a merged slice of time ranges of completed
transfers and adding those to the current transfers.
In a28239f005 we made --files-from obey --no-traverse. In the
process this caused --files-from without --no-traverse to do a
complete recursive scan unecessarily.
This was only noticeable in users of fs/march, so sync/copy/move/etc
not in ls/lsf/etc.
This fix makes sure that we use conventional directory listings in
fs/march unless `--files-from` and `--no-traverse` is set or
`--fast-list` is active.
Fixes#3619
It was reported that v1.49.4 which was accidentally compiled with
go1.13 instead of go1.12 produced errors like this:
Failed to get StartPageToken: Get https://www.googleapis.com/drive/v3/changes/startPageToken?XXX: stream error: stream ID 1789; INTERNAL_ERROR
IO error: open file failed: Get https://www.googleapis.com/drive/v3/files/XXX?alt=media: stream error: stream ID 1781; INTERNAL_ERROR
These are errors from the http2 library. It appears that go1.13 when
communicating with google drive defaults to http2 whereas with go1.12
it doesn't.
It is unclear what is causing these errors, but retrying them since
they don't happen very often seems like a valid strategy.
This was fixed in v1.49.5 by compiling with go1.12 - this fix is
designed to work with go1.13
See: https://forum.rclone.org/t/1-49-4-plex-internal-errors-on-google-drive/12108/
Before this fix, attempting to set a non top level environment
variable would fail with "Couldn't find flag".
This fixes it by passing in the flags that the env var is being set
from.
Fixes#3615
Before this change --update would transfer any file which was newer
than the destination regardless of whether it had changed or not.
This is needlessly wasteful of bandwidth.
After this change --update will only transfer files if they are newer
**and** they are different (checked with checksum and size).
'Couldn't find file - Need to Transfer' changed to 'Need to transfer -
File Not Found at Destination' because while reading the debug logs, it
confuses with failure in operation.
changes:
- chunker: remove GetTier and SetTier
- remove wdmrcompat metaformat
- remove fastopen strategy
- make hash_type option non-advanced
- adverise hash support when possible
- add metadata field "ver", run strict checks
- describe internal behavior in comments
- improve documentation
note:
wdmrcompat used to write file name in the metadata, so maximum metadata
size was 1K; removing it allows to cap size by 200 bytes now.
