minio/cmd/background-heal-ops.go
Anis Elleuch 456b2ef6eb
Avoid healing to be stuck with many concurrent event listeners (#10111)
If there are many listeners to bucket notifications or to the trace
subsystem, healing fails to work properly since it suspends itself when
the number of concurrent connections is above a certain threshold.

These connections are also continuous and not costly (*no disk access*),
it is okay to just ignore them in waitForLowHTTPReq().
2020-07-22 13:16:55 -07:00

169 lines
4.7 KiB
Go

/*
* MinIO Cloud Storage, (C) 2019 MinIO, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package cmd
import (
"context"
"path"
"time"
"github.com/minio/minio/cmd/logger"
"github.com/minio/minio/pkg/madmin"
)
// healTask represents what to heal along with options
// path: '/' => Heal disk formats along with metadata
// path: 'bucket/' or '/bucket/' => Heal bucket
// path: 'bucket/object' => Heal object
type healTask struct {
bucket string
object string
versionID string
opts madmin.HealOpts
// Healing response will be sent here
responseCh chan healResult
}
// healResult represents a healing result with a possible error
type healResult struct {
result madmin.HealResultItem
err error
}
// healRoutine receives heal tasks, to heal buckets, objects and format.json
type healRoutine struct {
tasks chan healTask
doneCh chan struct{}
}
// Add a new task in the tasks queue
func (h *healRoutine) queueHealTask(task healTask) {
h.tasks <- task
}
func waitForLowHTTPReq(tolerance int32) {
// Bucket notification and http trace are not costly, it is okay to ignore them
// while counting the number of concurrent connections
tolerance += int32(globalHTTPListen.NumSubscribers() + globalHTTPTrace.NumSubscribers())
if httpServer := newHTTPServerFn(); httpServer != nil {
// Wait at max 10 minute for an inprogress request before proceeding to heal
waitCount := 600
// Any requests in progress, delay the heal.
for (httpServer.GetRequestCount() >= tolerance) &&
waitCount > 0 {
waitCount--
time.Sleep(1 * time.Second)
}
}
}
// Wait for heal requests and process them
func (h *healRoutine) run(ctx context.Context, objAPI ObjectLayer) {
for {
select {
case task, ok := <-h.tasks:
if !ok {
break
}
// Wait and proceed if there are active requests
waitForLowHTTPReq(int32(globalEndpoints.NEndpoints()))
var res madmin.HealResultItem
var err error
switch {
case task.bucket == nopHeal:
continue
case task.bucket == SlashSeparator:
res, err = healDiskFormat(ctx, objAPI, task.opts)
case task.bucket != "" && task.object == "":
res, err = objAPI.HealBucket(ctx, task.bucket, task.opts.DryRun, task.opts.Remove)
case task.bucket != "" && task.object != "":
res, err = objAPI.HealObject(ctx, task.bucket, task.object, task.versionID, task.opts)
}
if task.bucket != "" && task.object != "" {
ObjectPathUpdated(path.Join(task.bucket, task.object))
}
task.responseCh <- healResult{result: res, err: err}
case <-h.doneCh:
return
case <-ctx.Done():
return
}
}
}
func newHealRoutine() *healRoutine {
return &healRoutine{
tasks: make(chan healTask),
doneCh: make(chan struct{}),
}
}
func initBackgroundHealing(ctx context.Context, objAPI ObjectLayer) {
// Run the background healer
globalBackgroundHealRoutine = newHealRoutine()
go globalBackgroundHealRoutine.run(ctx, objAPI)
nh := newBgHealSequence()
// Heal any disk format and metadata early, if possible.
if err := nh.healDiskMeta(); err != nil {
if newObjectLayerFn() != nil {
// log only in situations, when object layer
// has fully initialized.
logger.LogIf(nh.ctx, err)
}
}
globalBackgroundHealState.LaunchNewHealSequence(nh)
}
// healDiskFormat - heals format.json, return value indicates if a
// failure error occurred.
func healDiskFormat(ctx context.Context, objAPI ObjectLayer, opts madmin.HealOpts) (madmin.HealResultItem, error) {
res, err := objAPI.HealFormat(ctx, opts.DryRun)
// return any error, ignore error returned when disks have
// already healed.
if err != nil && err != errNoHealRequired {
return madmin.HealResultItem{}, err
}
// Healing succeeded notify the peers to reload format and re-initialize disks.
// We will not notify peers if healing is not required.
if err == nil {
// Notify servers in background and retry if needed.
go func() {
retry:
for _, nerr := range globalNotificationSys.ReloadFormat(opts.DryRun) {
if nerr.Err != nil {
if nerr.Err.Error() == errServerNotInitialized.Error() {
time.Sleep(time.Second)
goto retry
}
logger.GetReqInfo(ctx).SetTags("peerAddress", nerr.Host.String())
logger.LogIf(ctx, nerr.Err)
}
}
}()
}
return res, nil
}