Profiler API to record time spent in queues#4653
Conversation
9a08d83 to
52e9549
Compare
richardstartin
changed the title
e030299 to
03bbf09
Compare
There was a problem hiding this comment.
this could stay at 32 if, for example, the reference to AgentTrace were removed. We can get away with just 32 bits for the timestamp because we know the queuing duration will be less than 25 days.
There was a problem hiding this comment.
How would you make Continuation work without an AgentTrace instance?
For the record, I replace it by a Context instance in my PoC.
There was a problem hiding this comment.
there's a reference chain from the span to the trace, so a couple of indirections can be traded for a reduction in space.
ed46ed5 to
612dd50
Compare
612dd50 to
42f1a80
Compare
There was a problem hiding this comment.
How would you make Continuation work without an AgentTrace instance?
For the record, I replace it by a Context instance in my PoC.
There was a problem hiding this comment.
Any implicit reason to call ScopeManager.continueSpan() before ScopeManager.recordContinuationLifeSpan()?
(by implicit, I mean I could not find one but I expect to missed it 😓 )
There was a problem hiding this comment.
yes, to ensure the scope is on the stack (which has the side effect of activating it) - this probably requires a comment, so thanks for picking up on it.
There was a problem hiding this comment.
as discussed offline, I think we can wait for changes in this area and keep this PR scoped to the profiler changes
49ab276 to
f00787c
Compare
richardstartin
changed the title
f00787c to
436956f
Compare
richardstartin
added
the
tag: no release notes
3 participants
What Does This Do
Motivation
Additional Notes