feat: add debouncing to provisionerd rpc calls #5198
Conversation
f786491 to
5fc0c10
Compare
3993cec to
e3d6f9b
Compare
Seems to be BE only so dismissing just to be safe it's not accidentally merged
| @@ -205,8 +211,8 @@ func (p *Server) connect(ctx context.Context) { | |||
| if p.isClosed() { | |||
| return | |||
| } | |||
| ticker := time.NewTicker(p.opts.PollInterval) | |||
| defer ticker.Stop() | |||
| timer := time.NewTimer(p.opts.JobPollInterval) | |||
There was a problem hiding this comment.
Wouldn't it be better to only apply this delay if it failed to acquire a job? This means we will acquire them as quick as possible without delays if there are a lot of jobs queued, but during periods of low jobs we will only attempt to acquire a job every 5 seconds.
There was a problem hiding this comment.
I guess in theory yeah, but it's currently architected in a way where this is still ticking while the provisioner is running a job, so it'd require a bit of a refactor for this to work.
There was a problem hiding this comment.
couldn't you just do something like:
for {
ok :=acquireJob()
if !ok {
select {
case <-ctx.Done():
return ctx.Err()
case <-time.After(5 * time.Second):
}
}or something? time.After's only leak if they haven't fired yet, so since the interval is small and the only time this for loop will exit is if the app is shutting down it seems fine here
| lastAcquireMutex.RLock() | ||
| if !lastAcquire.IsZero() && time.Since(lastAcquire) < p.opts.JobPollDebounce { | ||
| lastAcquireMutex.RUnlock() | ||
| return | ||
| } | ||
| lastAcquireMutex.RUnlock() |
There was a problem hiding this comment.
Should disable this debounce if running in tests
There was a problem hiding this comment.
It's only enabled in coder server
coderd2is running these changes