Documentation ¶
Overview ¶
Package mpather provides per-mountpath concepts.
- Copyright (c) 2018-2024, NVIDIA CORPORATION. All rights reserved.
Package mpather provides per-mountpath concepts.
- Copyright (c) 2018-2021, NVIDIA CORPORATION. All rights reserved.
Index ¶
Constants ¶
View Source
const ( ThrottleMinDur = time.Millisecond ThrottleAvgDur = time.Millisecond * 10 ThrottleMaxDur = time.Millisecond * 100 )
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Jgroup ¶ added in v1.3.16
type Jgroup struct {
// contains filtered or unexported fields
}
Jgroup runs jogger per mountpath which walk the entire bucket and call callback on each of the encountered object. When jogger encounters error it stops and informs other joggers about the error (so they stop too).
func NewJoggerGroup ¶
func (*Jgroup) ListenFinished ¶ added in v1.3.16
func (jg *Jgroup) ListenFinished() <-chan struct{}
type JgroupOpts ¶ added in v1.3.16
type JgroupOpts struct { VisitObj func(lom *core.LOM, buf []byte) error VisitCT func(ct *core.CT, buf []byte) error Slab *memsys.Slab Bck cmn.Bck Buckets cmn.Bcks Prefix string CTs []string DoLoad LoadType // if specified, lom.Load(lock type) Parallel int // num parallel calls IncludeCopy bool // visit copies (aka replicas) PerBucket bool // num joggers = (num mountpaths) x (num buckets) SkipGloballyMisplaced bool // skip globally misplaced Throttle bool // true: pace itself depending on disk utilization // contains filtered or unexported fields }
type WorkerGroup ¶
type WorkerGroup struct {
// contains filtered or unexported fields
}
WorkerGroup starts one worker per mountpath; each worker receives (*core.LOM) tasks and executes the specified callback.
func NewWorkerGroup ¶
func NewWorkerGroup(opts *WorkerGroupOpts) *WorkerGroup
func (*WorkerGroup) PostLIF ¶ added in v1.3.19
func (wg *WorkerGroup) PostLIF(lom *core.LOM) (chanFull bool, err error)
func (*WorkerGroup) Run ¶
func (wg *WorkerGroup) Run()
func (*WorkerGroup) Stop ¶
func (wg *WorkerGroup) Stop() (n int)
Stop aborts all the workers. It should be called after we are sure no more new tasks will be dispatched.
Click to show internal directories.
Click to hide internal directories.