maintenance register currently records the maintenance repo exclusively
within the user's global configuration, but other configuration files
may be relevant when running maintenance if they are included from the
global config. This option allows the user to choose where maintenance
repos are recorded.
Signed-off-by: Ronan Pigott <ronan@rjp.ie>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
The 'git maintenance unregister' subcommand has a step that removes the
current repository from the multi-valued maitenance.repo config key.
This fails if the repository is not listed in that key. This makes
running 'git maintenance unregister' twice result in a failure in the
second instance.
This failure exit code is helpful, but its message is not. Add a new
die() message that explicitly calls out the failure due to the
repository not being registered.
In some cases, users may want to run 'git maintenance unregister' just
to make sure that background jobs will not start on this repository, but
they do not want to check to see if it is registered first. Add a new
'--force' option that will siltently succeed if the repository is not
already registered.
Also add an extra test of 'git maintenance unregister' at a point where
there are no registered repositories. This should fail without --force.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a CONFIGURATION section to the documentation of various built-ins,
for those cases where the relevant config/NAME.txt describes
configuration that is only used by the relevant built-in documented in
git-NAME.txt. Subsequent commits will handle more complex cases.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Reviewed-by: Matheus Tavares <matheus.bernardino@usp.br>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The synopsis for 'git maintenance' did not include the commands other
than the 'run' command. Update this to include the others. The 'start'
command is the only one of these that parses additional options, and
then only the --scheduler option.
Also move the 'register' command down after 'stop' and before
'unregister' for a logical grouping of the commands instead of an
alphabetical one. The diff makes it look as three other commands are
moved up.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The existing mechanism for scheduling background maintenance is done
through cron. On Linux systems managed by systemd, systemd provides an
alternative to schedule recurring tasks: systemd timers.
The main motivations to implement systemd timers in addition to cron
are:
* cron is optional and Linux systems running systemd might not have it
installed.
* The execution of `crontab -l` can tell us if cron is installed but not
if the daemon is actually running.
* With systemd, each service is run in its own cgroup and its logs are
tagged by the service inside journald. With cron, all scheduled tasks
are running in the cron daemon cgroup and all the logs of the
user-scheduled tasks are pretended to belong to the system cron
service.
Concretely, a user that doesn’t have access to the system logs won’t
have access to the log of their own tasks scheduled by cron whereas
they will have access to the log of their own tasks scheduled by
systemd timer.
Although `cron` attempts to send email, that email may go unseen by
the user because these days, local mailboxes are not heavily used
anymore.
In order to schedule git maintenance, we need two unit template files:
* ~/.config/systemd/user/git-maintenance@.service
to define the command to be started by systemd and
* ~/.config/systemd/user/git-maintenance@.timer
to define the schedule at which the command should be run.
Those units are templates that are parameterized by the frequency.
Based on those templates, 3 timers are started:
* git-maintenance@hourly.timer
* git-maintenance@daily.timer
* git-maintenance@weekly.timer
The command launched by those three timers are the same as with the
other scheduling methods:
/path/to/git for-each-repo --exec-path=/path/to
--config=maintenance.repo maintenance run --schedule=%i
with the full path for git to ensure that the version of git launched
for the scheduled maintenance is the same as the one used to run
`maintenance start`.
The timer unit contains `Persistent=true` so that, if the computer is
powered down when a maintenance task should run, the task will be run
when the computer is back powered on.
Signed-off-by: Lénaïc Huard <lenaic@lhuard.fr>
Acked-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Depending on the system, different schedulers can be used to schedule
the hourly, daily and weekly executions of `git maintenance run`:
* `launchctl` for MacOS,
* `schtasks` for Windows and
* `crontab` for everything else.
`git maintenance run` now has an option to let the end-user explicitly
choose which scheduler he wants to use:
`--scheduler=auto|crontab|launchctl|schtasks`.
When `git maintenance start --scheduler=XXX` is run, it not only
registers `git maintenance run` tasks in the scheduler XXX, it also
removes the `git maintenance run` tasks from all the other schedulers to
ensure we cannot have two schedulers launching concurrent identical
tasks.
The default value is `auto` which chooses a suitable scheduler for the
system.
`git maintenance stop` doesn't have any `--scheduler` parameter because
this command will try to remove the `git maintenance run` tasks from all
the available schedulers.
Signed-off-by: Lénaïc Huard <lenaic@lhuard.fr>
Acked-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The 'prefetch' maintenance task previously forced the following refspec
for each remote:
+refs/heads/*:refs/prefetch/<remote>/*
If a user has specified a more strict refspec for the remote, then this
prefetch task downloads more objects than necessary.
The previous change introduced the '--prefetch' option to 'git fetch'
which manipulates the remote's refspec to place all resulting refs into
refs/prefetch/, with further partitioning based on the destinations of
those refspecs.
Update the documentation to be more generic about the destination refs.
Do not mention custom refspecs explicitly, as that does not need to be
highlighted in this documentation. The important part of placing refs in
refs/prefetch/ remains.
Reported-by: Tom Saeger <tom.saeger@oracle.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is valuable to collect loose refs into a more compressed form. This
is typically the packed-refs file, although this could be the reftable
in the future. Having packed refs can be extremely valuable in repos
with many tags or remote branches that are not modified by the local
user, but still are necessary for other queries.
For instance, with many exploded refs, commands such as
git describe --tags --exact-match HEAD
can be very slow (multiple seconds). This command in particular is used
by terminal prompts to show when a detatched HEAD is pointing to an
existing tag, so having it be slow causes significant delays for users.
Add a new 'pack-refs' maintenance task. It runs 'git pack-refs --all
--prune' to move loose refs into a packed form. For now, that is the
packed-refs file, but could adjust to other file formats in the future.
This is the first of several sub-tasks of the 'gc' task that could be
extracted to their own tasks. In this process, we should not change the
behavior of the 'gc' task since that remains the default way to keep
repositories maintained. Creating a new task for one of these sub-tasks
only provides more customization options for those choosing to not use
the 'gc' task. It is certainly possible to have both the 'gc' and
'pack-refs' tasks enabled and run regularly. While they may repeat
effort, they do not conflict in a destructive way.
The 'auto_condition' function pointer is left NULL for now. We could
extend this in the future to have a condition check if pack-refs should
be run during 'git maintenance run --auto'.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Reviewed-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Follow-up on the "maintenance part-3" which introduced scheduled
maintenance tasks to support platforms whose native scheduling
methods are not 'cron'.
* ds/maintenance-part-4:
maintenance: use Windows scheduled tasks
maintenance: use launchctl on macOS
maintenance: include 'cron' details in docs
maintenance: extract platform-specific scheduling
Git's background maintenance uses cron by default, but this is not
available on Windows. Instead, integrate with Task Scheduler.
Tasks can be scheduled using the 'schtasks' command. There are several
command-line options that can allow for some advanced scheduling, but
unfortunately these seem to all require authenticating using a password.
Instead, use the "/xml" option to pass an XML file that contains the
configuration for the necessary schedule. These XML files are based on
some that I exported after constructing a schedule in the Task Scheduler
GUI. These options only run background maintenance when the user is
logged in, and more fields are populated with the current username and
SID at run-time by 'schtasks'.
Since the GIT_TEST_MAINT_SCHEDULER environment variable allows us to
specify 'schtasks' as the scheduler, we can test the Windows-specific
logic on other platforms. Thus, add a check that the XML file written
by Git is valid when xmllint exists on the system.
Since we use a temporary file for the XML files sent to 'schtasks', we
prefix the random characters with the frequency so it is easier to
examine the proper file during tests. Instead of an exact match on the
'args' file, we 'grep' for the arguments other than the filename.
There is a deficiency in the current design. Windows has two kinds of
applications: GUI applications that start by "winmain()" and console
applications that start by "main()". Console applications are attached
to a new Console window if they are not already associated with a GUI
application. This means that every hour the scheudled task launches a
command window for the scheduled tasks. Not only is this visually
obtrusive, but it also takes focus from whatever else the user is
doing!
A simple fix would be to insert a GUI application that acts as a shim
between the scheduled task and Git. This is currently possible in Git
for Windows by setting the <Command> tag equal to
C:\Program Files\Git\git-bash.exe
with options "--hide --no-needs-console --command=cmd\git.exe"
followed by the arguments currently used. Since git-bash.exe is not
included in Windows builds of core Git, I chose to leave out this
feature. My plan is to submit a small patch to Git for Windows that
converts the use of git.exe with this use of git-bash.exe in the
short term. In the long term, we can consider creating this GUI
shim application within core Git, perhaps in contrib/.
Co-authored-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The existing mechanism for scheduling background maintenance is done
through cron. The 'crontab -e' command allows updating the schedule
while cron itself runs those commands. While this is technically
supported by macOS, it has some significant deficiencies:
1. Every run of 'crontab -e' must request elevated privileges through
the user interface. When running 'git maintenance start' from the
Terminal app, it presents a dialog box saying "Terminal.app would
like to administer your computer. Administration can include
modifying passwords, networking, and system settings." This is more
alarming than what we are hoping to achieve. If this alert had some
information about how "git" is trying to run "crontab" then we would
have some reason to believe that this dialog might be fine. However,
it also doesn't help that some scenarios just leave Git waiting for
a response without presenting anything to the user. I experienced
this when executing the command from a Bash terminal view inside
Visual Studio Code.
2. While cron initializes a user environment enough for "git config
--global --show-origin" to show the correct config file information,
it does not set up the environment enough for Git Credential Manager
Core to load credentials during a 'prefetch' task. My prefetches
against private repositories required re-authenticating through UI
pop-ups in a way that should not be required.
The solution is to switch from cron to the Apple-recommended [1]
'launchd' tool.
[1] https://developer.apple.com/library/archive/documentation/MacOSX/Conceptual/BPSystemStartup/Chapters/ScheduledJobs.html
The basics of this tool is that we need to create XML-formatted
"plist" files inside "~/Library/LaunchAgents/" and then use the
'launchctl' tool to make launchd aware of them. The plist files
include all of the scheduling information, along with the command-line
arguments split across an array of <string> tags.
For example, here is my plist file for the weekly scheduled tasks:
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0"><dict>
<key>Label</key><string>org.git-scm.git.weekly</string>
<key>ProgramArguments</key>
<array>
<string>/usr/local/libexec/git-core/git</string>
<string>--exec-path=/usr/local/libexec/git-core</string>
<string>for-each-repo</string>
<string>--config=maintenance.repo</string>
<string>maintenance</string>
<string>run</string>
<string>--schedule=weekly</string>
</array>
<key>StartCalendarInterval</key>
<array>
<dict>
<key>Day</key><integer>0</integer>
<key>Hour</key><integer>0</integer>
<key>Minute</key><integer>0</integer>
</dict>
</array>
</dict>
</plist>
The schedules for the daily and hourly tasks are more complicated
since we need to use an array for the StartCalendarInterval with
an entry for each of the six days other than the 0th day (to avoid
colliding with the weekly task), and each of the 23 hours other
than the 0th hour (to avoid colliding with the daily task).
The "Label" value is currently filled with "org.git-scm.git.X"
where X is the frequency. We need a different plist file for each
frequency.
The launchctl command needs to be aligned with a user id in order
to initialize the command environment. This must be done using
the 'launchctl bootstrap' subcommand. This subcommand is new as
of macOS 10.11, which was released in September 2015. Before that
release the 'launchctl load' subcommand was recommended. The best
source of information on this transition I have seen is available
at [2]. The current design does not preclude a future version that
detects the available fatures of 'launchctl' to use the older
commands. However, it is best to rely on the newest version since
Apple might completely remove the deprecated version on short
notice.
[2] https://babodee.wordpress.com/2016/04/09/launchctl-2-0-syntax/
To remove a schedule, we must run 'launchctl bootout' with a valid
plist file. We also need to 'bootout' a task before the 'bootstrap'
subcommand will succeed, if such a task already exists.
The need for a user id requires us to run 'id -u' which works on
POSIX systems but not Windows. Further, the need for fully-qualitifed
path names including $HOME behaves differently in the Git internals and
the external test suite. The $HOME variable starts with "C:\..." instead
of the "/c/..." that is provided by Git in these subcommands. The test
therefore has a prerequisite that we are not on Windows. The cross-
platform logic still allows us to test the macOS logic on a Linux
machine.
We can verify the commands that were run by 'git maintenance start'
and 'git maintenance stop' by injecting a script that writes the
command-line arguments into GIT_TEST_MAINT_SCHEDULER.
An earlier version of this patch accidentally had an opening
"<dict>" tag when it should have had a closing "</dict>" tag. This
was caught during manual testing with actual 'launchctl' commands,
but we do not want to update developers' tasks when running tests.
It appears that macOS includes the "xmllint" tool which can verify
the XML format. This is useful for any system that might contain
the tool, so use it whenever it is available.
We strive to make these tests work on all platforms, but Windows caused
some headaches. In particular, the value of getuid() called by the C
code is not guaranteed to be the same as `$(id -u)` invoked by a test.
This is because `git.exe` is a native Windows program, whereas the
utility programs run by the test script mostly utilize the MSYS2 runtime,
which emulates a POSIX-like environment. Since the purpose of the test
is to check that the input to the hook is well-formed, the actual user
ID is immaterial, thus we can work around the problem by making the the
test UID-agnostic. Another subtle issue is the $HOME environment
variable being a Windows-style path instead of a Unix-style path. We can
be more flexible here instead of expecting exact path matches.
Helped-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Co-authored-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a missing "a" before "bunch".
Signed-off-by: Martin Ågren <martin.agren@gmail.com>
Acked-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Advanced and expert users may want to know how 'git maintenance start'
schedules background maintenance in order to customize their own
schedules beyond what the maintenance.* config values allow. Start a new
set of sections in git-maintenance.txt that describe how 'cron' is used
to run these tasks.
This is particularly valuable for users who want to inspect what Git is
doing or for users who want to customize the schedule further. Having a
baseline can provide a way forward for users who have never worked with
cron schedules.
Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The 'git maintenance run' subcommand takes a lock on the object database
to prevent concurrent processes from competing for resources. This is an
important safety measure to prevent possible repository corruption and
data loss.
This feature can lead to confusing behavior if a user is not aware of
it. Add a TROUBLESHOOTING section to the 'git maintenance' builtin
documentation that discusses these tradeoffs. The short version of this
section is that Git will not corrupt your repository, but if the list of
scheduled tasks takes longer than an hour then some scheduled tasks may
be dropped due to this object database collision. For example, a
long-running "daily" task at midnight might prevent an "hourly" task
from running at 1AM.
The opposite is also possible, but less likely as long as the "hourly"
tasks are much faster than the "daily" and "weekly" tasks.
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The 'git maintenance (register|start)' subcommands add the current
repository to the global Git config so maintenance will operate on that
repository. It does not specify what maintenance should occur or how
often.
To make it simple for users to start background maintenance with a
recommended schedlue, update the 'maintenance.strategy' config option in
both the 'register' and 'start' subcommands. This allows users to
customize beyond the defaults using individual
'maintenance.<task>.schedule' options, but also the user can opt-out of
this strategy using 'maintenance.strategy=none'.
Helped-by: Martin Ågren <martin.agren@gmail.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add new subcommands to 'git maintenance' that start or stop background
maintenance using 'cron', when available. This integration is as simple
as I could make it, barring some implementation complications.
The schedule is laid out as follows:
0 1-23 * * * $cmd maintenance run --schedule=hourly
0 0 * * 1-6 $cmd maintenance run --schedule=daily
0 0 * * 0 $cmd maintenance run --schedule=weekly
where $cmd is a properly-qualified 'git for-each-repo' execution:
$cmd=$path/git --exec-path=$path for-each-repo --config=maintenance.repo
where $path points to the location of the Git executable running 'git
maintenance start'. This is critical for systems with multiple versions
of Git. Specifically, macOS has a system version at '/usr/bin/git' while
the version that users can install resides at '/usr/local/bin/git'
(symlinked to '/usr/local/libexec/git-core/git'). This will also use
your locally-built version if you build and run this in your development
environment without installing first.
This conditional schedule avoids having cron launch multiple 'git
for-each-repo' commands in parallel. Such parallel commands would likely
lead to the 'hourly' and 'daily' tasks competing over the object
database lock. This could lead to to some tasks never being run! Since
the --schedule=<frequency> argument will run all tasks with _at least_
the given frequency, the daily runs will also run the hourly tasks.
Similarly, the weekly runs will also run the daily and hourly tasks.
The GIT_TEST_CRONTAB environment variable is not intended for users to
edit, but instead as a way to mock the 'crontab [-l]' command. This
variable is set in test-lib.sh to avoid a future test from accidentally
running anything with the cron integration from modifying the user's
schedule. We use GIT_TEST_CRONTAB='test-tool crontab <file>' in our
tests to check how the schedule is modified in 'git maintenance
(start|stop)' commands.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In preparation for launching background maintenance from the 'git
maintenance' builtin, create register/unregister subcommands. These
commands update the new 'maintenance.repos' config option in the global
config so the background maintenance job knows which repositories to
maintain.
These commands allow users to add a repository to the background
maintenance list without disrupting the actual maintenance mechanism.
For example, a user can run 'git maintenance register' when no
background maintenance is running and it will not start the background
maintenance. A later update to start running background maintenance will
then pick up this repository automatically.
The opposite example is that a user can run 'git maintenance unregister'
to remove the current repository from background maintenance without
halting maintenance for other repositories.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Maintenance currently triggers when certain data-size thresholds are
met, such as number of pack-files or loose objects. Users may want to
run certain maintenance tasks based on frequency instead. For example,
a user may want to perform a 'prefetch' task every hour, or 'gc' task
every day. To help these users, update the 'git maintenance run' command
to include a '--schedule=<frequency>' option. The allowed frequencies
are 'hourly', 'daily', and 'weekly'. These values are also allowed in a
new config value 'maintenance.<task>.schedule'.
The 'git maintenance run --schedule=<frequency>' checks the '*.schedule'
config value for each enabled task to see if the configured frequency is
at least as frequent as the frequency from the '--schedule' argument. We
use the following order, for full clarity:
'hourly' > 'daily' > 'weekly'
Use new 'enum schedule_priority' to track these values numerically.
The following cron table would run the scheduled tasks with the correct
frequencies:
0 1-23 * * * git -C <repo> maintenance run --schedule=hourly
0 0 * * 1-6 git -C <repo> maintenance run --schedule=daily
0 0 * * 0 git -C <repo> maintenance run --schedule=weekly
This cron schedule will run --schedule=hourly every hour except at
midnight. This avoids a concurrent run with the --schedule=daily that
runs at midnight every day except the first day of the week. This avoids
a concurrent run with the --schedule=weekly that runs at midnight on
the first day of the week. Since --schedule=daily also runs the
'hourly' tasks and --schedule=weekly runs the 'hourly' and 'daily'
tasks, we will still see all tasks run with the proper frequencies.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous change cleaned up loose objects using the
'loose-objects' that can be run safely in the background. Add a
similar job that performs similar cleanups for pack-files.
One issue with running 'git repack' is that it is designed to
repack all pack-files into a single pack-file. While this is the
most space-efficient way to store object data, it is not time or
memory efficient. This becomes extremely important if the repo is
so large that a user struggles to store two copies of the pack on
their disk.
Instead, perform an "incremental" repack by collecting a few small
pack-files into a new pack-file. The multi-pack-index facilitates
this process ever since 'git multi-pack-index expire' was added in
19575c7 (multi-pack-index: implement 'expire' subcommand,
2019-06-10) and 'git multi-pack-index repack' was added in ce1e4a1
(midx: implement midx_repack(), 2019-06-10).
The 'incremental-repack' task runs the following steps:
1. 'git multi-pack-index write' creates a multi-pack-index file if
one did not exist, and otherwise will update the multi-pack-index
with any new pack-files that appeared since the last write. This
is particularly relevant with the background fetch job.
When the multi-pack-index sees two copies of the same object, it
stores the offset data into the newer pack-file. This means that
some old pack-files could become "unreferenced" which I will use
to mean "a pack-file that is in the pack-file list of the
multi-pack-index but none of the objects in the multi-pack-index
reference a location inside that pack-file."
2. 'git multi-pack-index expire' deletes any unreferenced pack-files
and updaes the multi-pack-index to drop those pack-files from the
list. This is safe to do as concurrent Git processes will see the
multi-pack-index and not open those packs when looking for object
contents. (Similar to the 'loose-objects' job, there are some Git
commands that open pack-files regardless of the multi-pack-index,
but they are rarely used. Further, a user that self-selects to
use background operations would likely refrain from using those
commands.)
3. 'git multi-pack-index repack --bacth-size=<size>' collects a set
of pack-files that are listed in the multi-pack-index and creates
a new pack-file containing the objects whose offsets are listed
by the multi-pack-index to be in those objects. The set of pack-
files is selected greedily by sorting the pack-files by modified
time and adding a pack-file to the set if its "expected size" is
smaller than the batch size until the total expected size of the
selected pack-files is at least the batch size. The "expected
size" is calculated by taking the size of the pack-file divided
by the number of objects in the pack-file and multiplied by the
number of objects from the multi-pack-index with offset in that
pack-file. The expected size approximates how much data from that
pack-file will contribute to the resulting pack-file size. The
intention is that the resulting pack-file will be close in size
to the provided batch size.
The next run of the incremental-repack task will delete these
repacked pack-files during the 'expire' step.
In this version, the batch size is set to "0" which ignores the
size restrictions when selecting the pack-files. It instead
selects all pack-files and repacks all packed objects into a
single pack-file. This will be updated in the next change, but
it requires doing some calculations that are better isolated to
a separate change.
These steps are based on a similar background maintenance step in
Scalar (and VFS for Git) [1]. This was incredibly effective for
users of the Windows OS repository. After using the same VFS for Git
repository for over a year, some users had _thousands_ of pack-files
that combined to up to 250 GB of data. We noticed a few users were
running into the open file descriptor limits (due in part to a bug
in the multi-pack-index fixed by af96fe3 (midx: add packs to
packed_git linked list, 2019-04-29).
These pack-files were mostly small since they contained the commits
and trees that were pushed to the origin in a given hour. The GVFS
protocol includes a "prefetch" step that asks for pre-computed pack-
files containing commits and trees by timestamp. These pack-files
were grouped into "daily" pack-files once a day for up to 30 days.
If a user did not request prefetch packs for over 30 days, then they
would get the entire history of commits and trees in a new, large
pack-file. This led to a large number of pack-files that had poor
delta compression.
By running this pack-file maintenance step once per day, these repos
with thousands of packs spanning 200+ GB dropped to dozens of pack-
files spanning 30-50 GB. This was done all without removing objects
from the system and using a constant batch size of two gigabytes.
Once the work was done to reduce the pack-files to small sizes, the
batch size of two gigabytes means that not every run triggers a
repack operation, so the following run will not expire a pack-file.
This has kept these repos in a "clean" state.
[1] https://github.com/microsoft/scalar/blob/master/Scalar.Common/Maintenance/PackfileMaintenanceStep.cs
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
One goal of background maintenance jobs is to allow a user to
disable auto-gc (gc.auto=0) but keep their repository in a clean
state. Without any cleanup, loose objects will clutter the object
database and slow operations. In addition, the loose objects will
take up extra space because they are not stored with deltas against
similar objects.
Create a 'loose-objects' task for the 'git maintenance run' command.
This helps clean up loose objects without disrupting concurrent Git
commands using the following sequence of events:
1. Run 'git prune-packed' to delete any loose objects that exist
in a pack-file. Concurrent commands will prefer the packed
version of the object to the loose version. (Of course, there
are exceptions for commands that specifically care about the
location of an object. These are rare for a user to run on
purpose, and we hope a user that has selected background
maintenance will not be trying to do foreground maintenance.)
2. Run 'git pack-objects' on a batch of loose objects. These
objects are grouped by scanning the loose object directories in
lexicographic order until listing all loose objects -or-
reaching 50,000 objects. This is more than enough if the loose
objects are created only by a user doing normal development.
We noticed users with _millions_ of loose objects because VFS
for Git downloads blobs on-demand when a file read operation
requires populating a virtual file.
This step is based on a similar step in Scalar [1] and VFS for Git.
[1] https://github.com/microsoft/scalar/blob/master/Scalar.Common/Maintenance/LooseObjectsStep.cs
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When working with very large repositories, an incremental 'git fetch'
command can download a large amount of data. If there are many other
users pushing to a common repo, then this data can rival the initial
pack-file size of a 'git clone' of a medium-size repo.
Users may want to keep the data on their local repos as close as
possible to the data on the remote repos by fetching periodically in
the background. This can break up a large daily fetch into several
smaller hourly fetches.
The task is called "prefetch" because it is work done in advance
of a foreground fetch to make that 'git fetch' command much faster.
However, if we simply ran 'git fetch <remote>' in the background,
then the user running a foreground 'git fetch <remote>' would lose
some important feedback when a new branch appears or an existing
branch updates. This is especially true if a remote branch is
force-updated and this isn't noticed by the user because it occurred
in the background. Further, the functionality of 'git push
--force-with-lease' becomes suspect.
When running 'git fetch <remote> <options>' in the background, use
the following options for careful updating:
1. --no-tags prevents getting a new tag when a user wants to see
the new tags appear in their foreground fetches.
2. --refmap= removes the configured refspec which usually updates
refs/remotes/<remote>/* with the refs advertised by the remote.
While this looks confusing, this was documented and tested by
b40a50264a (fetch: document and test --refmap="", 2020-01-21),
including this sentence in the documentation:
Providing an empty `<refspec>` to the `--refmap` option
causes Git to ignore the configured refspecs and rely
entirely on the refspecs supplied as command-line arguments.
3. By adding a new refspec "+refs/heads/*:refs/prefetch/<remote>/*"
we can ensure that we actually load the new values somewhere in
our refspace while not updating refs/heads or refs/remotes. By
storing these refs here, the commit-graph job will update the
commit-graph with the commits from these hidden refs.
4. --prune will delete the refs/prefetch/<remote> refs that no
longer appear on the remote.
5. --no-write-fetch-head prevents updating FETCH_HEAD.
We've been using this step as a critical background job in Scalar
[1] (and VFS for Git). This solved a pain point that was showing up
in user reports: fetching was a pain! Users do not like waiting to
download the data that was created while they were away from their
machines. After implementing background fetch, the foreground fetch
commands sped up significantly because they mostly just update refs
and download a small amount of new data. The effect is especially
dramatic when paried with --no-show-forced-udpates (through
fetch.showForcedUpdates=false).
[1] https://github.com/microsoft/scalar/blob/master/Scalar.Common/Maintenance/FetchStep.cs
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, a normal run of "git maintenance run" will only run the 'gc'
task, as it is the only one enabled. This is mostly for backwards-
compatible reasons since "git maintenance run --auto" commands replaced
previous "git gc --auto" commands after some Git processes. Users could
manually run specific maintenance tasks by calling "git maintenance run
--task=<task>" directly.
Allow users to customize which steps are run automatically using config.
The 'maintenance.<task>.enabled' option then can turn on these other
tasks (or turn off the 'gc' task).
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A user may want to only run certain maintenance tasks in a certain
order. Add the --task=<task> option, which allows a user to specify an
ordered list of tasks to run. These cannot be run multiple times,
however.
Here is where our array of maintenance_task pointers becomes critical.
We can sort the array of pointers based on the task order, but we do not
want to move the struct data itself in order to preserve the hashmap
references. We use the hashmap to match the --task=<task> arguments into
the task struct data.
Keep in mind that the 'enabled' member of the maintenance_task struct is
a placeholder for a future 'maintenance.<task>.enabled' config option.
Thus, we use the 'enabled' member to specify which tasks are run when
the user does not specify any --task=<task> arguments. The 'enabled'
member should be ignored if --task=<task> appears.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The first new task in the 'git maintenance' builtin is the
'commit-graph' task. This updates the commit-graph file
incrementally with the command
git commit-graph write --reachable --split
By writing an incremental commit-graph file using the "--split"
option we minimize the disruption from this operation. The default
behavior is to merge layers until the new "top" layer is less than
half the size of the layer below. This provides quick writes most
of the time, with the longer writes following a power law
distribution.
Most importantly, concurrent Git processes only look at the
commit-graph-chain file for a very short amount of time, so they
will verly likely not be holding a handle to the file when we try
to replace it. (This only matters on Windows.)
If a concurrent process reads the old commit-graph-chain file, but
our job expires some of the .graph files before they can be read,
then those processes will see a warning message (but not fail).
This could be avoided by a future update to use the --expire-time
argument when writing the commit-graph.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Maintenance activities are commonly used as steps in larger scripts.
Providing a '--quiet' option allows those scripts to be less noisy when
run on a terminal window. Turn this mode on by default when stderr is
not a terminal.
Pipe the option to the 'git gc' child process.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The 'gc' builtin is our current entrypoint for automatically maintaining
a repository. This one tool does many operations, such as repacking the
repository, packing refs, and rewriting the commit-graph file. The name
implies it performs "garbage collection" which means several different
things, and some users may not want to use this operation that rewrites
the entire object database.
Create a new 'maintenance' builtin that will become a more general-
purpose command. To start, it will only support the 'run' subcommand,
but will later expand to add subcommands for scheduling maintenance in
the background.
For now, the 'maintenance' builtin is a thin shim over the 'gc' builtin.
In fact, the only option is the '--auto' toggle, which is handed
directly to the 'gc' builtin. The current change is isolated to this
simple operation to prevent more interesting logic from being lost in
all of the boilerplate of adding a new builtin.
Use existing builtin/gc.c file because we want to share code between the
two builtins. It is possible that we will have 'maintenance' replace the
'gc' builtin entirely at some point, leaving 'git gc' as an alias for
some specific arguments to 'git maintenance run'.
Create a new test_subcommand helper that allows us to test if a certain
subcommand was run. It requires storing the GIT_TRACE2_EVENT logs in a
file. A negation mode is available that will be used in later tests.
Helped-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren
Ronan Pigott
Derrick Stolee
Ævar Arnfjörð Bjarmason
Lénaïc Huard
Derrick Stolee
Junio C Hamano
Thomas Ackermann
Martin Ågren