housekeeping

package
v16.10.2 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Apr 9, 2024 License: MIT Imports: 20 Imported by: 0

Documentation

Index

Constants

View Source
const (

	// ReferenceLockfileGracePeriod is the grace period when cleaning up lock files of individual references in the
	// repository. Lock files existing less than this period are ignored.
	ReferenceLockfileGracePeriod = 1 * time.Hour
	// ReftableLockfileGracePeriod is the grace period when cleaning up lock files reftable.
	ReftableLockfileGracePeriod = 1 * time.Hour
)
View Source
const (
	// RepackObjectsStrategyIncrementalWithUnreachable performs an incremental repack by writing
	// all loose objects into a new packfile, regardless of their reachability. The loose
	// objects will be deleted.
	RepackObjectsStrategyIncrementalWithUnreachable = RepackObjectsStrategy("incremental_with_unreachable")
	// RepackObjectsStrategyFullWithCruft performs a full repack by writing all reachable
	// objects into a new packfile. Unreachable objects will be written into a separate cruft
	// packfile.
	RepackObjectsStrategyFullWithCruft = RepackObjectsStrategy("full_with_cruft")
	// RepackObjectsStrategyFullWithUnreachable performs a full repack by writing all reachable
	// objects into a new packfile. Packed unreachable objects will be appended to the packfile
	// and redundant loose object files will be deleted.
	//
	// Note that this will not include unreachable loose objects, but only packed loose objects.
	// git-repack(1) does not currently expose an option to make it include all objects.
	// Combined with geometric repacks though this is acceptable as the geometric strategy will
	// include all loose objects.
	RepackObjectsStrategyFullWithUnreachable = RepackObjectsStrategy("full_with_unreachable")
	// RepackObjectsStrategyGeometric performs an geometric repack. This strategy will repack
	// packfiles so that the resulting pack structure forms a geometric sequence in the number
	// of objects. Loose objects will get soaked up as part of the repack regardless of their
	// reachability.
	RepackObjectsStrategyGeometric = RepackObjectsStrategy("geometric")
)
View Source
const (
	// WorktreesPrefix is a subdirectory in the repository where Gitaly creates worktrees.
	WorktreesPrefix = "worktrees"
	// GitlabWorktreePrefix is a subdirectory in the repository where Gitaly creates worktrees.
	GitlabWorktreePrefix = "gitlab-worktree"
)
View Source
const (
	// FullRepackCooldownPeriod is the cooldown period that needs to pass since the last full
	// repack before we consider doing another full repack.
	FullRepackCooldownPeriod = 5 * 24 * time.Hour
)
View Source
const (
	// LooseObjectLimit is the limit of loose objects we accept both when doing incremental
	// repacks and when pruning objects.
	LooseObjectLimit = 1024
)

Variables

This section is empty.

Functions

func CleanupWorktrees

func CleanupWorktrees(ctx context.Context, repo *localrepo.Repo) error

CleanupWorktrees cleans up stale and disconnected worktrees for the given repository.

func FindBrokenLooseReferences added in v16.10.0

func FindBrokenLooseReferences(ctx context.Context, repoPath string) ([]string, error)

FindBrokenLooseReferences return the list of broken refs. A ref is considered to be broken when the loose ref file is empty. We don't validate the existence of referred objects.

func FindPackedRefsLock added in v16.10.0

func FindPackedRefsLock(ctx context.Context, repoPath string) ([]string, error)

FindPackedRefsLock returns stale lockfiles for the packed-refs file.

func FindPackedRefsNew added in v16.10.0

func FindPackedRefsNew(ctx context.Context, repoPath string) ([]string, error)

FindPackedRefsNew returns stale temporary packed-refs files.

func FindServerInfo added in v16.10.0

func FindServerInfo(ctx context.Context, repoPath string) ([]string, error)

FindServerInfo returns files generated by git-update-server-info(1). These files are only required to serve Git fetches via the dumb HTTP protocol, which we don't serve at all. It's thus safe to remove all of those files without a grace period.

func FindStaleLockfiles added in v16.10.0

func FindStaleLockfiles(ctx context.Context, repoPath string) ([]string, error)

FindStaleLockfiles finds a subset of lockfiles which may be created by git commands. We're quite conservative with what we're removing, we certainly don't just scan the repo for `*.lock` files. Instead, we only remove a known set of lockfiles which have caused problems in the past.

func FindStalePackFileLocks added in v16.10.0

func FindStalePackFileLocks(ctx context.Context, repoPath string) ([]string, error)

FindStalePackFileLocks finds packfile locks (`.keep` files) that git-receive-pack(1) and git-fetch-pack(1) write in order to not have the newly written packfile be deleted while refs have not yet been updated to point to their objects. Normally, these locks would get removed by both Git commands once the refs have been updated. But when the command gets killed meanwhile, then it can happen that the locks are left behind. As Git will never delete packfiles with an associated `.keep` file, the end result is that we may accumulate more and more of these locked packfiles over time.

func FindStaleReftableLock added in v16.10.0

func FindStaleReftableLock(_ context.Context, repoPath string) ([]string, error)

FindStaleReftableLock provides a function which scans the reftable folder for stale a reftable lock and deletes it.

func FindTemporaryObjects added in v16.10.0

func FindTemporaryObjects(ctx context.Context, repoPath string) ([]string, error)

FindTemporaryObjects find temporary objects of a repository. Those objects are created by some git processes and are sure to be removed when the processes are done.

func GetRepackGitConfig

func GetRepackGitConfig(ctx context.Context, repo storage.Repository, bitmap bool) []git.ConfigPair

GetRepackGitConfig returns configuration suitable for Git commands which write new packfiles.

func PerformFullRepackingWithUnreachable added in v16.9.0

func PerformFullRepackingWithUnreachable(ctx context.Context, repo *localrepo.Repo, cfg RepackObjectsConfig) error

PerformFullRepackingWithUnreachable performs a full repacking task using git-repack(1) command. This will omit packing objects part of alternates.

func PerformGeometricRepacking added in v16.9.0

func PerformGeometricRepacking(ctx context.Context, repo *localrepo.Repo, cfg RepackObjectsConfig) error

PerformGeometricRepacking performs geometric repacking task using git-repack(1) command. It allows us to merge multiple packfiles without having to rewrite all packfiles into one. This new "geometric" strategy tries to ensure that existing packfiles in the repository form a geometric sequence where each successive packfile contains at least n times as many objects as the preceding packfile. If the sequence isn't maintained, Git will determine a slice of packfiles that it must repack to maintain the sequence again. With this process, we can limit the number of packfiles that exist in the repository without having to repack all objects into a single packfile regularly. This repacking does not take reachability into account. For more information, https://about.gitlab.com/blog/2023/11/02/rearchitecting-git-object-database-mainentance-for-scale/#geometric-repacking

func PerformRepack added in v16.9.0

func PerformRepack(ctx context.Context, repo *localrepo.Repo, cfg RepackObjectsConfig, opts ...git.Option) error

PerformRepack performs `git-repack(1)` command on a repository with some pre-built configs.

func PruneEmptyConfigSections added in v16.10.0

func PruneEmptyConfigSections(ctx context.Context, repo *localrepo.Repo) (_ int, returnedErr error)

PruneEmptyConfigSections prunes all empty sections from the repo's config.

func PruneObjects

func PruneObjects(ctx context.Context, repo *localrepo.Repo, cfg PruneObjectsConfig) error

PruneObjects prunes loose objects from the repository that are already packed or which are unreachable and older than the configured expiry date.

func RemoveRefEmptyDirs added in v16.10.0

func RemoveRefEmptyDirs(ctx context.Context, repository *localrepo.Repo) (int, error)

RemoveRefEmptyDirs removes empty directories inside refs directory. They are left-over of ref deletion operations.

func RepackObjects

func RepackObjects(ctx context.Context, repo *localrepo.Repo, cfg RepackObjectsConfig) error

RepackObjects repacks objects in the given repository and updates the commit-graph. The way objects are repacked is determined via the RepackObjectsConfig.

func ValidateRepacking added in v16.9.0

func ValidateRepacking(cfg RepackObjectsConfig) (bool, error)

ValidateRepacking validates the input repacking config. This function any validating error and if the configuration is for full repack.

func WriteCommitGraph

func WriteCommitGraph(ctx context.Context, repo *localrepo.Repo, cfg WriteCommitGraphConfig) error

WriteCommitGraph updates the commit-graph in the given repository. The commit-graph is updated incrementally, except in the case where it doesn't exist yet or in case it is detected that the commit-graph is missing bloom filters.

Types

type CleanStaleDataConfig

type CleanStaleDataConfig struct {
	// StaleFileFinders contains the list of finder functions to find stale files in the repository.
	StaleFileFinders map[string]FindStaleFileFunc
	// RepoCleanups contains the list of clean-up functions for the repository.
	RepoCleanups map[string]CleanupRepoFunc
	// RepoCleanupWithTxManagers contains the list of clean-up functions with transaction manager support.
	RepoCleanupWithTxManagers map[string]cleanupRepoWithTxManagerFunc
}

CleanStaleDataConfig is the configuration for running CleanStaleData. It is used to define the different types of cleanups we want to run.

func DefaultStaleDataCleanup

func DefaultStaleDataCleanup() CleanStaleDataConfig

DefaultStaleDataCleanup is the default configuration for CleanStaleData which contains all the cleanup functions.

func OnlyStaleReferenceLockCleanup

func OnlyStaleReferenceLockCleanup(gracePeriod time.Duration) CleanStaleDataConfig

OnlyStaleReferenceLockCleanup returns a config which only contains a stale reference lock cleaner with the provided grace period.

type CleanupRepoFunc added in v16.10.0

type CleanupRepoFunc func(context.Context, *localrepo.Repo) (int, error)

CleanupRepoFunc is a type of function that performs a clean up task in the repository.

type EagerOptimizationStrategy

type EagerOptimizationStrategy struct {
	// contains filtered or unexported fields
}

EagerOptimizationStrategy is a strategy that will eagerly perform optimizations. All of the data structures will be optimized regardless of whether they already are in an optimal state or not.

func NewEagerOptimizationStrategy

func NewEagerOptimizationStrategy(info stats.RepositoryInfo) EagerOptimizationStrategy

NewEagerOptimizationStrategy creates a new EagerOptimizationStrategy.

func (EagerOptimizationStrategy) ShouldPruneObjects

ShouldPruneObjects always instructs the caller to prune objects, unless the repository is an object pool.

func (EagerOptimizationStrategy) ShouldRepackObjects

func (s EagerOptimizationStrategy) ShouldRepackObjects(ctx context.Context) (bool, RepackObjectsConfig)

ShouldRepackObjects always instructs the caller to repack objects. The strategy will always be to repack all objects into a single packfile. The bitmap will be written in case the repository does not have any alternates.

func (EagerOptimizationStrategy) ShouldRepackReferences

func (s EagerOptimizationStrategy) ShouldRepackReferences(context.Context) bool

ShouldRepackReferences always instructs the caller to repack references.

func (EagerOptimizationStrategy) ShouldWriteCommitGraph

ShouldWriteCommitGraph always instructs the caller to write the commit-graph. The strategy will always be to completely rewrite the commit-graph chain.

type FindStaleFileFunc added in v16.10.0

type FindStaleFileFunc func(context.Context, string) ([]string, error)

FindStaleFileFunc is a type of function that returns the list of stale files in the repository.

func FindStaleReferenceLocks added in v16.10.0

func FindStaleReferenceLocks(gracePeriod time.Duration) FindStaleFileFunc

FindStaleReferenceLocks provides a function which scans the refdb for stale locks and loose references against the provided grace period.

type HeuristicalOptimizationStrategy

type HeuristicalOptimizationStrategy struct {
	// contains filtered or unexported fields
}

HeuristicalOptimizationStrategy is an optimization strategy that is based on a set of heuristics.

func NewHeuristicalOptimizationStrategy

func NewHeuristicalOptimizationStrategy(gitVersion git.Version, info stats.RepositoryInfo) HeuristicalOptimizationStrategy

NewHeuristicalOptimizationStrategy constructs a heuristicalOptimizationStrategy for the given repository info. It derives all data from the repository so that the heuristics used by this repository can be decided without further disk reads.

func (HeuristicalOptimizationStrategy) ShouldPruneObjects

ShouldPruneObjects determines whether the repository has stale objects that should be pruned. Object pools are never pruned to not lose data in them, but otherwise we prune when we've found enough stale objects that might in fact get pruned.

func (HeuristicalOptimizationStrategy) ShouldRepackObjects

ShouldRepackObjects checks whether the repository's objects need to be repacked. This uses a set of heuristics that scales with the size of the object database: the larger the repository, the less frequent does it get a full repack.

func (HeuristicalOptimizationStrategy) ShouldRepackReferences

func (s HeuristicalOptimizationStrategy) ShouldRepackReferences(context.Context) bool

ShouldRepackReferences determines whether the repository's references need to be repacked based on heuristics. The more references there are, the more loose referencos may exist until they are packed again.

func (HeuristicalOptimizationStrategy) ShouldWriteCommitGraph

ShouldWriteCommitGraph determines whether we need to write the commit-graph and how it should be written.

type OptimizationStrategy

type OptimizationStrategy interface {
	// ShouldRepackObjects determines whether the repository needs to be repacked and, if so,
	// how it should be done.
	ShouldRepackObjects(context.Context) (bool, RepackObjectsConfig)
	// ShouldPruneObjects determines whether the repository has stale objects that should be
	// pruned and, if so, how it should be done.
	ShouldPruneObjects(context.Context) (bool, PruneObjectsConfig)
	// ShouldRepackReferences determines whether the repository's references need to be
	// repacked.
	ShouldRepackReferences(context.Context) bool
	// ShouldWriteCommitGraph determines whether we need to write the commit-graph and how it
	// should be written.
	ShouldWriteCommitGraph(context.Context) (bool, WriteCommitGraphConfig, error)
}

OptimizationStrategy is an interface to determine which parts of a repository should be optimized.

type PruneObjectsConfig

type PruneObjectsConfig struct {
	// ExpireBefore controls the grace period after which unreachable objects shall be pruned.
	// An unreachable object must be older than the given date in order to be considered for
	// deletion.
	ExpireBefore time.Time
}

PruneObjectsConfig determines which objects should be pruned in PruneObjects.

type RepackObjectsConfig

type RepackObjectsConfig struct {
	// Strategy determines the strategy with which to repack objects.
	Strategy RepackObjectsStrategy
	// WriteBitmap determines whether reachability bitmaps should be written or not. There is no
	// reason to set this to `false`, except for legacy compatibility reasons with existing RPC
	// behaviour
	WriteBitmap bool
	// WriteMultiPackIndex determines whether a multi-pack index should be written or not.
	WriteMultiPackIndex bool
	// CruftExpireBefore determines the cutoff date before which unreachable cruft objects shall
	// be expired and thus deleted.
	CruftExpireBefore time.Time
}

RepackObjectsConfig is configuration for RepackObjects.

type RepackObjectsStrategy

type RepackObjectsStrategy string

RepackObjectsStrategy defines how objects shall be repacked.

type WriteCommitGraphConfig

type WriteCommitGraphConfig struct {
	// ReplaceChain causes WriteCommitGraph to rewrite the complete commit-graph chain. This is
	// a lot more expensive than the default, incremental update of the commit-graph chains but
	// may be required in certain cases to fix up commit-graphs.
	ReplaceChain bool
}

WriteCommitGraphConfig contains configuration that can be passed to WriteCommitGraph to alter its default behaviour.

func WriteCommitGraphConfigForRepository

func WriteCommitGraphConfigForRepository(ctx context.Context, repo *localrepo.Repo) (WriteCommitGraphConfig, error)

WriteCommitGraphConfigForRepository returns the optimal default-configuration for the given repo. By default, the configuration will ask for an incremental commit-graph update. If the preexisting commit-graph is missing bloom filters though then the whole commit-graph chain will be rewritten.

Directories

Path Synopsis

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL