Fix for span leaks

[Mixaster995]

Signed-off-by: Mikhail Avramenko avramenkomihail15@gmail.com

Description

Added another span logger for 'infinite' Find case scenario(registry)

Issue link

#1085

How Has This Been Tested?

• Added unit testing to cover
• Tested manually
• Tested by integration testing
• Have not tested

Types of changes

• Bug fix
• New functionallity
• Documentation
• Refactoring
• CI

[denis-tingaikin]

Please let's get rid of static variables to use struct fields.

Suggested change

	var loggers []*childSpanLogger
	var ticker = time.NewTicker(time.Minute)
	var once sync.Once
	var lock sync.Mutex

[Mixaster995]

ok, done

[denis-tingaikin]

Please let's try to re-use exist span logger to avoid code duplication.

Suggested change

	type childSpanLogger struct {
	span opentracing.Span
	childSpan opentracing.Span
	entries map[interface{}]interface{}
	lock sync.RWMutex
	operation string
	}
	type todoSpanLogger struct {
	*originalSpanLogger
	}

[Mixaster995]

ok, done

[denis-tingaikin]

Please let's try to use cond or fanout pattern for subscriptions on events and use shared state.

Suggested change

	go func() {
	for {
	select {
	case <-ticker.C:
	for _, l := range loggers {
	if l.span != nil {
	l.lock.Lock()
	l.childSpan.Finish()
	l.childSpan = l.span.Tracer().StartSpan(l.operation, opentracing.ChildOf(l.span.Context()))
	l.lock.Unlock()
	}
	}
	case <-ctx.Done():
	return
	}
	}
	}()

[Mixaster995]

ok, done

[edwarnicke]

@Mixaster995 Is this still a good PR in need of review?

[Mixaster995]

@edwarnicke no, it was a test solution for reported problem, but the problem itself is not clear and really hard to reproduce or catch. Thus, we decided to postpone investigation of this problem until better times.

[edwarnicke]

@Mixaster995 Got it