opensource/podman
1
0
Fork 0
mirror of https://github.com/containers/podman.git synced 2025-05-17 23:26:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

work queue: simplify and use a wait group

Browse Source 
Simplify the work-queue implementation by using a wait group. Once all
queued work items are done, the channel can be closed.

The system tests revealed a flake (i.e., #14351) which indicated that
the service container does not always get stopped which suggests a race
condition when queuing items.  Those items are queued in a goroutine to
prevent potential dead locks if the queue ever filled up too quickly.
The race condition in question is that if a work item queues another,
the goroutine for queuing may not be scheduled fast enough and the
runtime shuts down; it seems to happen fairly easily on the slow CI
machines.  The wait group fixes this race and allows for simplifying
the code.

Also increase the queue's buffer size to 10 to make things slightly
faster.

[NO NEW TESTS NEEDED] as we are fixing a flake.

Fixes: #14351
Signed-off-by: Valentin Rothberg <vrothberg@redhat.com>
This commit is contained in:
Valentin Rothberg
2022-05-25 10:08:17 +02:00
parent b13184dfb4
commit 4a447a2133
2 changed files with 11 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
libpod/runtime.go
View File

@ -11,6 +11,7 @@ import (
"regexp"
"strconv"
"strings"
"sync"
"syscall"
"time"
@ -87,8 +88,8 @@ type Runtime struct {
lockManager lock.Manager
// Worker
workerShutdown chan bool
workerChannel chan func()
workerChannel chan func()
workerGroup sync.WaitGroup
// syslog describes whenever logrus should log to the syslog as well.
// Note that the syslog hook will be enabled early in cmd/podman/syslog_linux.go
@ -823,12 +824,9 @@ func (r *Runtime) Shutdown(force bool) error {
return define.ErrRuntimeStopped
}
if r.workerShutdown != nil {
// Signal the worker routine to shutdown. The routine will
// process all pending work items and then read from the
// channel; we're blocked until all work items have been
// processed.
r.workerShutdown <- true
if r.workerChannel != nil {
r.workerGroup.Wait()
close(r.workerChannel)
}
r.valid = false

33
libpod/runtime_worker.go
View File

@ -1,40 +1,17 @@
package libpod
import (
"time"
)
func (r *Runtime) startWorker() {
if r.workerChannel == nil {
r.workerChannel = make(chan func(), 1)
r.workerShutdown = make(chan bool)
}
r.workerChannel = make(chan func(), 10)
go func() {
for {
// Make sure to read all workers before
// checking if we're about to shutdown.
for len(r.workerChannel) > 0 {
w := <-r.workerChannel
w()
}
select {
// We'll read from the shutdown channel only when all
// items above have been processed.
//
// (*Runtime).Shutdown() will block until until the
// item is read.
case <-r.workerShutdown:
return
default:
time.Sleep(100 * time.Millisecond)
}
for w := range r.workerChannel {
w()
r.workerGroup.Done()
}
}()
}
func (r *Runtime) queueWork(f func()) {
r.workerGroup.Add(1)
go func() {
r.workerChannel <- f
}()