Use separate stages for download and upload. In the common case where
it downloads the file from one remote and then uploads to the other,
those are by far the most expensive operations, and there's a decent
chance the two remotes bottleneck on different resources.
Suppose it's being run with -J2 and a bunch of 10 mb files. Two threads
will be started both downloading from the src remote. They will probably
finish at the same time. Then two threads will be started uploading to
the dst remote. They will probably take the same time as well. Before
this change, it would alternate back and forth, bottlenecking on src and dst.
With this change, as soon as the two threads start uploading to dst, two
more threads are able to start, downloading from src. So bandwidth to
both remotes is saturated more often.
Other commands that use transferStages only send in one direction at a
time. So the worker threads for the other direction will sit idle, and
there will be no change in their behavior.
Sponsored-by: Dartmouth College's DANDI project
Lock the local content for drop after getting it from src, to prevent another
process from using the local content as a copy and dropping it from src,
which would prevent dropping the local content after sending it to dest.
Support resuming an interrupted move that downloaded the content from
src, leaving the local content populated. In this case, the location log
has not been updated to say the content is present locally, so we can
assume that it's resuming and go ahead and drop the local content after
sending it to dest.
Note that if a `git-annex get` is being ran at the same time as a
`git-annex move --from --to`, it may get a file just before the move
processes it. So the location log has not been updated yet, and the move
thinks it's resuming. Resulting in local copy being dropped after it's
sent to the dest. This race is something we'll just have to live with,
it seems.
I also gave up on the idea of checking if the location log had been updated
by a `git-annex get` that is ran at the same time. That wouldn't work, because
the location log is precached in the seek stage, so reading it again after
sending the content to dest would not notice changes made to it, unless the cache
were invalidated, which would slow it down a lot. That idea anyway was subject
to races where it would not detect the concurrent `git-annex get`.
So concurrent `git-annex get` will have results that may be surprising.
To make that less surprising, updated the documentation of this feature to
be explicit that it downloads content to the local repository
temporarily.
Sponsored-by: Dartmouth College's DANDI project
Prep for move --to --from, which needs to download from a src repo
without updating the location log for the local repo, before sending the
content on to the dest repo.
Note that caller of download' already update the log themselves.
See previous commit a422a056f2
that pushed it up to download from getViaTmpFrom.
(Also removed in passing a debug print + readline that I accidentially
committed last week on this branch.)
Sponsored-by: Dartmouth College's DANDI project
Note that when this is specified and an older git-annex is used to
enableremote such a special remote, it will simply ignore the cost= field
and use whatever the default cost is.
In passing, fixed adb to support the remote.name.cost and
remote.name.cost-command configs.
Sponsored-by: Dartmouth College's DANDI project
readish ignores a trailing string after a number, but to support values
like "YYYY:MM:DD" which it makes sense to compare lexographically,
require the whole string to be parsed as a number in order to enable
numeric comparison.
Sponsored-by: Max Thoursie on Patreon
Improve handling of some .git/annex/ subdirectories being on other
filesystems, in the bittorrent special remote, and youtube-dl integration,
and git-annex addurl.
The only one of these that I've confirmed to be a problem is in the
bittorrent special remote when .git/annex/tmp and .git/annex/othertmp are
on different filesystems.
As well as auditing for renameFile, also audited for createLink, all of
those are ok as are the other remaining renameFile calls. Also audited all
code paths that use .git/annex/othertmp, and did not find any other
cross-device problems. So, removing mention of othertmp needing to be on
the same device.
Sponsored-by: Dartmouth College's Datalad project
Change --metadata comparisons < > <= and >= to fall back to lexicographical
comparisons when one or both values being compared are not numbers.
Sponsored-by: Erik Bjäreholt on Patreon
Fix a hang that occasionally occurred during commands such as move.
(A bug introduced in 10.20220927, in
commit 6a3bd283b8)
The restage.log was kept locked while running a complex index refresh
action. In an unusual situation, that action could need to write to the
restage log, which caused a deadlock.
The solution is a two-stage process. First the restage.log is moved to a
work file, which is done with the lock held. Then the content of the work
file is read and processed, which happens without the lock being held.
This is all done in a crash-safe manner.
Note that streamRestageLog may not be fully safe to run concurrently
with itself. That's ok, because restagePointerFiles uses it with the
index lock held, so only one can be run at a time.
streamRestageLog does delete the restage.old file at the end without
locking. If a calcRestageLog is run concurrently, it will either see the
file content before it was deleted, or will see it's missing. Either is
ok, because at most this will cause calcRestageLog to report more
work remains to be done than there is.
Sponsored-by: Dartmouth College's Datalad project
Debian is going to drop youtube-dl which is not active upstream, and yt-dlp
is the replacement. This will make it be used if youtube-dl gets removed.
If an old version of youtube-dl remains installed, git-annex will still use
it. That might not be desirable, but changing git-annex to use yt-dlp in
preference to youtube-dl when both are installed risks breaking when
the user has annex.youtube-dl-options set to something that is supported
by youtube-dl, but not by yt-dlp.
Sponsored-by: Boyd Stephen Smith Jr. on Patreon
init: Avoid scanning for annexed files, which can be lengthy in a
large repository. Instead that scan is done on demand. This lets git-annex
init be run and some query commands be used in a repository without
waiting.
Note that autoinit already behaved this way, so while this will mean some
commands like git-annex get/unlock/add will do the scan the first time run,
that is not really a significant behavior change.
And, it's really better to have a consistent behavior. The reason for
the inconsistency was a strange bug discussed in
b3c4579c79. Avoiding reconcileStaged in
init will keep avoiding whatever that was.
Sponsored-by: Dartmouth College's DANDI project
Clean the standalone environment before running the su command
to run "sh". Otherwise, PATH leaked through, causing it to run
git-annex.linux/bin/sh, but GIT_ANNEX_DIR was not set,
which caused that script to not work:
[2022-10-26 15:07:02.145466106] (Utility.Process) process [938146] call: pkexec ["sh","-c","cd '/home/joey/tmp/git-annex.linux/r' && '/home/joey/tmp/git-annex.linux/git-annex' 'enable-tor' '1000'"]
/home/joey/tmp/git-annex.linux/bin/sh: 4: exec: /exe/sh: not found
Changed programPath to not use GIT_ANNEX_PROGRAMPATH,
but instead run the scripts at the top of GIT_ANNEX_DIR.
That works both when the standalone environment is set up, and when it's
not.
Sponsored-by: Kevin Mueller on Patreon
Make --batch mode handle unstaged annexed files consistently whether the
file is unlocked or not. Before this, a unstaged locked file
would have the symlink on disk examined and operated on in --batch mode,
while an unstaged unlocked file would be skipped.
Note that, when not in batch mode, unstaged files are skipped over too.
That is actually somewhat new behavior; as late as 7.20191114 a
command like `git-annex whereis .` would operate on unstaged locked
files and skip over unstaged unlocked files. That changed during
optimisation of CmdLine.Seek with apparently little fanfare or notice.
Turns out that rmurl still behaved that way when given an unstaged file
on the command line. It was changed to use lookupKeyStaged to
handle its --batch mode. That also affected its non-batch mode, but
since that's just catching up to the change earlier made to most
other commands, I have not mentioed that in the changelog.
It may be that other uses of lookupKey should also change to
lookupKeyStaged. But it may also be that would slow down some things,
or lead to unwanted behavior changes, so I've kept the changes minimal
for now.
An example of a place where the use of lookupKey is better than
lookupKeyStaged is in Command.AddUrl, where it looks to see if the file
already exists, and adds the url to the file when so. It does not matter
there whether the file is staged or not (when it's locked). The use of
lookupKey in Command.Unused likewise seems good (and faster).
Sponsored-by: Nicholas Golder-Manning on Patreon
When running eg git-annex get, for each file it has to read from and
write to the keys database. But it's reading exclusively from one table,
and writing to a different table. So, it is not necessary to flush the
write to the database before reading. This avoids writing the database
once per file, instead it will buffer 1000 changes before writing.
Benchmarking getting 1000 small files from a local origin,
git-annex get now takes 13.62s, down from 22.41s!
git-annex drop now takes 9.07s, down from 18.63s!
Wowowowowowowow!
(It would perhaps have been better if there were separate databases for
the two tables. At least it would have avoided this complexity. Ah well,
this is better than splitting the table in a annex.version upgrade.)
Sponsored-by: Dartmouth College's Datalad project
The flush was only done Annex.run' to make sure that the queue was flushed
before git-annex exits. But, doing it there means that as soon as one
change gets queued, it gets flushed soon after, which contributes to
excessive writes to the database, slowing git-annex down.
(This does not yet speed git-annex up, but it is a stepping stone to
doing so.)
Database queues do not autoflush when garbage collected, so have to
be flushed explicitly. I don't think it's possible to make them
autoflush (except perhaps if git-annex sqitched to using ResourceT..).
The comment in Database.Keys.closeDb used to be accurate, since the
automatic flushing did mean that all writes reached the database even
when closeDb was not called. But now, closeDb or flushDb needs to be
called before stopping using an Annex state. So, removed that comment.
In Remote.Git, change to using quiesce everywhere that it used to use
stopCoProcesses. This means that uses on onLocal in there are just as
slow as before. I considered only calling closeDb on the local git remotes
when git-annex exits. But, the reason that Remote.Git calls stopCoProcesses
in each onLocal is so as not to leave git processes running that have files
open on the remote repo, when it's on removable media. So, it seemed to make
sense to also closeDb after each one, since sqlite may also keep files
open. Although that has not seemed to cause problems with removable
media so far. It was also just easier to quiesce in each onLocal than
once at the end. This does likely leave performance on the floor, so
could be revisited.
In Annex.Content.saveState, there was no reason to close the db,
flushing it is enough.
The rest of the changes are from auditing for Annex.new, and making
sure that quiesce is called, after any action that might possibly need
it.
After that audit, I'm pretty sure that the change to Annex.run' is
safe. The only concern might be that this does let more changes get
queued for write to the db, and if git-annex is interrupted, those will be
lost. But interrupting git-annex can obviously already prevent it from
writing the most recent change to the db, so it must recover from such
lost data... right?
Sponsored-by: Dartmouth College's Datalad project
When importing from versioned remotes, fix tracking of the content of
deleted files.
Only S3 supports versioning so far, so only it was affected.
But, the draft import/export interface for external remotes also seemed to
need a change, so that versionedExport could be set.
Well, actually, fix a typo that has always been in the implementation of
that. "inbacked" used to work, but let's not tell users about that; they
might try to use it and expect git-annex to keep supporting the typo..
Sponsored-by: Jack Hill on Patreon
The user sets these hooks deliberately so they should always be run. For
example this allows hooks to be used to manage file permissions on NTFS
volumes in WSL1.
This is much easier and less failure-prone than having the user run
git update-index --refresh themselves.
Sponsored-by: Dartmouth College's DANDI project
When pointer files need to be restaged, they're first written to the
log, and then when the restage operation runs, it reads the log. This
way, if the git-annex process is interrupted before it can do the
restaging, a later git-annex process can do it.
Currently, this lets a git-annex get/drop command be interrupted and
then re-ran, and as long as it gets/drops additional files, it will
clean up after the interrupted command. But more changes are
needed to make it easier to restage after an interrupted process.
Kept using the git queue to run the restage action, even though the
list of files that it builds up for that action is not actually used by
the action. This could perhaps be simplified to make restaging a cleanup
action that gets registered, rather than using the git queue for it. But
I wasn't sure if that would cause visible behavior changes, when eg
dropping a large number of files, currently the git queue flushes
periodically, and so it restages incrementally, rather than all at the
end.
In restagePointerFiles, it reads the restage log twice, once to get
the number of files and size, and a second time to process it.
This seemed better than reading the whole file into memory, since
potentially a huge number of files could be in there. Probably the OS
will cache the file in memory and there will not be much performance
impact. It might be better to keep running tallies in another file
though. But updating that atomically with the log seems hard.
Also note that it's possible for calcRestageLog to see a different file
than streamRestageLog does. More files may be added to the log in
between. That is ok, it will only cause the filterprocessfaster heuristic to
operate with slightly out of date information, so it may make the wrong
choice for the files that got added and be a little slower than ideal.
Sponsored-by: Dartmouth College's DANDI project
Fixes updating git index file after getting an unlocked file when
annex.stalldetection is set.
The transferrer may want to send additional protocol messages when it's
shut down. Closing the read handle prevented it from doing that, and caused
it to crash rather than cleanly shutting down.
Draining the handle without processing the protocol seemed ok to do,
because anything it outputs is going to be some side message displayed
at shutdown. Displaying those once per transferrer process that is running
seems unncessary.
Sponsored-by: Dartmouth College's DANDI project
This partly fixes an issue where there are duplicate files in the
special remote, and the first file gets swapped with another duplicate,
or deleted. The swap case is fixed by this, the deleted case will need
other changes.
This makes retrieveExportWithContentIdentifier take a list of allowed
ContentIdentifier, same as storeExportWithContentIdentifier,
removeExportWithContentIdentifier, and
checkPresentExportWithContentIdentifier.
Of the special remotes that support importtree, borg is a special case
and does not use content identifiers, S3 I assume can't get mixed up
like this, directory certainly has the problem, and adb also appears to
have had the problem.
Sponsored-by: Graham Spencer on Patreon
Use curl for downloads from git remotes when annex.url-options and other
git configs are set.
If the url needs a password, curl will fail, and git credential will not be
used to prompt for it. But the user can set --netrc in url-options and
put the password in the netrc file.
This also means that url-options settings like -4 will take effect.
That was the case before commit 1883f7ef8f
forced conduit to be used.
autoEnableSpecialRemotes runs a subprocess, and if the uuid for a git
remote has not been probed yet, that will do a http get that will prompt
for a password. And then the parent process will subsequently prompt
for a password when getting annexed files from the remote.
So the solution is for autoEnableSpecialRemotes to run remoteList before
the subprocess, which will probe for the uuid for the git remote in the
same process that will later be used to get annexed files.
But, Remote.Git imports Annex.Init, and Remote.List imports Remote.Git,
so Annex.Init cannot import Remote.List. Had to pass remoteList into
functions in Annex.Init to get around this dependency loop.
When accessing a git remote over http needs a git credential prompt for a
password, cache it for the lifetime of the git-annex process, rather than
repeatedly prompting.
The git-lfs special remote already caches the credential when discovering
the endpoint. And presumably commands like git pull do as well, since they
may download multiple urls from a remote.
The TMVar CredentialCache is read, so two concurrent calls to
getBasicAuthFromCredential will both prompt for a credential.
There would already be two concurrent password prompts in such a case,
and existing uses of `prompt` probably avoid it. Anyway, it's no worse
than before.
Help the user get annex.dbdir configured when their filesystem is not
one that sqlite works on.
The change in Database.Handle makes an error from sqlite not be ignored
besides being displayed, which it was before. I can't see any reason
git-annex would want to ignore these errors.
I chose to use the fsck database rather than the keys database because
opening the keys database populates it, and see commit
b3c4579c79.
The placement of the call to checkSqliteWorks inside checkInitializeAllowed
avoids annex.uuid getting set before it's called.
Sponsored-by: Dartmouth College's Datalad project
Use curl when annex.security.allowed-url-schemes includes an url scheme not
supported by git-annex internally, as long as
annex.security.allowed-ip-addresses is configured to allow using curl.
Sponsored-by: Luke Shumaker on Patreon
This allows annex.dbdir to be set globally or always set to the same
value when needed. Each repository uses a subdirectory of it.
Sponsored-by: Dartmouth College's Datalad project
Completes work started in e60766543f
I've verified that all the sqlite databases get stored in annex.dbdir
and are created successfully. If annex.dbdir does not exist, it will be
created; its parent directory must already exist though.
Sponsored-by: Dartmouth College's Datalad project
This should not change the behavior of it, unless there are multiple top
directories, and then it should behave the same as if there was a single
top directory that was actually above the directory to be created.
Sponsored-by: Dartmouth College's Datalad project
WIP: This is mostly complete, but there is a problem: createDirectoryUnder
throws an error when annex.dbdir is set to outside the git repo.
annex.dbdir is a workaround for filesystems where sqlite does not work,
due to eg, the filesystem not properly supporting locking.
It's intended to be set before initializing the repository. Changing it
in an existing repository can be done, but would be the same as making a
new repository and moving all the annexed objects into it. While the
databases get recreated from the git-annex branch in that situation, any
information that is in the databases but not stored in the branch gets
lost. It may be that no information ever gets stored in the databases
that cannot be reconstructed from the branch, but I have not verified
that.
Sponsored-by: Dartmouth College's Datalad project
Work around bug in git 2.37 that causes a segfault when when
core.untrackedCache is set, and broke git-annex init.
Depending on when git gets fixed and how widely the buggy versions are
used, this could be reverted quite soon, or need to linger for a long time.
It only makes git-annex init a tiny bit slower in a new repo.
Sponsored-by: Max Thoursie on Patreon
(And v9 later on to v10.)
When v9/v10 were added, making v8 automatically upgrade was deferred
"for a few months" to prevent interoperability problems if users also
have an old version of git-annex. Of course that could still be the
case, but there has been a good amount of time and this can't be put off
forever.
Allow setting annex.autoupgraderepository to false to avoid this upgrade.
Previously, that only prevented upgrades from no longer supported git-annex
versions, but v8 is still supported, and users may want to keep on v8 to
interoperate with an old git-annex version.
Sponsored-by: Boyd Stephen Smith Jr. on Patreon
I was thinking that discardIncompleteAppend would make it strict, since
it looks at the end of the bytestring. But, it's applied lazily..
This probably fixes windows, which was failing:
git-annex.exe: .git\annex\journal\trust.log: DeleteFile "\\\\?\\C:\\Users\\runneradmin\\.t\\5\\tmprepo22\\.git\\annex\\journal\\trust.log": permission denied (The process cannot access the file because it is being used by another process.)
I would like for a new repo version to enable appends, but to do so
safely would need a v11 followed by a 1 year delay followed by a v12
that does it. Since a similar v9 and v10 transition is currently
happening, and is less than 6 months along in most repos, it does not
feel wise to stack up another year-long transition behind that. What if
I need to hurry up a new repo version for some other change?
Added todo so I remember to make this change at some time when a v11
and probably v12 repo version do make sense.
Sponsored-by: Dartmouth College's DANDI project
An append that is interrupted and writes part of a line is now dealt
with by subsequent reads and appends. This also handles a read that
happens at the same time as an append to the file.
Old versions of git-annex will still see a partially written line,
and could get confused. Since appends are currently done for url logs
and location logs, the confusion is limited to a substring of the actual
url or UUID of the remote being read. This will not affect writes, since
the journal file is locked when reading in preparation for writing.
However, the bad data can be output by git-annex and used by other
things, or could cause surprising behavior by git-annex. Including eg,
downloading the content of the wrong url.
So, something needs to be done to prevent old versions of git-annex from
running in a repository where this appending is being done..
Sponsored-by: Dartmouth College's DANDI project
This reverts commit 2e6e9876e3.
This is gonna be needed after all.. The append will only be atomic if
the journal is locked, because the file being appended will have to be
moved out of the way to avoid an old version of git-annex seeing an
incomplete write to it. When git-annex finds that the file is not in the
journal, and checks the append location, locking will be needed to avoid
a race causing it to miss it in the append location too due to it being
moved back to the journal.
Added annex.alwayscompact setting which can be unset to speed up writes to
the git-annex branch in some cases.
Sponsored-by: Dartmouth College's DANDI project
