You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The concurrent batch processor exposes its in-flight count, but we know that memory can pile up in the receiver waiting to be admitted by the concurrent batch processor (CBP). We'd benefit from being able to know how much memory is tied up in active requests that is stalled by the CBP's in-flight limit.
This functionality makes sense for all receivers; the community has discussed integrating memory limiter behavior as an extension, in which case the number of bytes in flight at the receiver would be used to fast-fail requests w/ RESOURCE_EXHAUSTED. For now, we'd like to track this quantity using an UpDownCounter.
Since items of telemetry vary in size, I think it would be useful in connection to know how many items are represented by the quantity of in-flight bytes. This suggests two new metrics, one for bytes and one for items, both counting in-flight numbers.
The text was updated successfully, but these errors were encountered:
… items pending (#150)
Resolves#148
This PR adds two metrics to track the number of items and bytes that are
pending in the receiver. Specifically when used with the
`concurrentbatchprocessor` requests can be have a long lifetime in
receiver due to the in flight byte limiter in the
`concurrentbatchprocessor`.
The concurrent batch processor exposes its in-flight count, but we know that memory can pile up in the receiver waiting to be admitted by the concurrent batch processor (CBP). We'd benefit from being able to know how much memory is tied up in active requests that is stalled by the CBP's in-flight limit.
This functionality makes sense for all receivers; the community has discussed integrating memory limiter behavior as an extension, in which case the number of bytes in flight at the receiver would be used to fast-fail requests w/ RESOURCE_EXHAUSTED. For now, we'd like to track this quantity using an UpDownCounter.
Since items of telemetry vary in size, I think it would be useful in connection to know how many items are represented by the quantity of in-flight bytes. This suggests two new metrics, one for bytes and one for items, both counting in-flight numbers.
The text was updated successfully, but these errors were encountered: