Re: need to write metrics/trace data to collector backend in fire and forget manner

[Sheng Wu]

I never work on the backend-js lib, but, I think
1. maxBufferSize doesn't affect the period of sending data.
2. I think gRPC has an onComplete signal that would be sent back from
the OAP server with an empty command object for every bulk.


Sheng Wu 吴晟
Twitter, wusheng1108

Pankaj Jain  于2023年8月9日周三 23:09写道：
>
> Hello Skywalking team,
>
> The Skywalking NodeJS agent is causing very high CPU utilization for NodeJS
> applications being instrumented. To lower down the CPU consumption, I am
> thinking
> to less frequently send the data and send those without waiting for the
> response using gRPC. So,
> 1. will try testing by increasing the maxBufferSize value from default 1000
> to higher no.
> 2. Assuming Async gRPC IO is happening to send the metric/span/signal data,
> is there a way I can ensure it is done in fire and forget manner; kind of
> not having callback functions to process the response. I understand this
> may not be reliable but we are ok with this.
>
> Any input/guidance is highly appreciated.
>
> Thanks -Pankaj

need to write metrics/trace data to collector backend in fire and forget manner

[Pankaj Jain]

Hello Skywalking team,

The Skywalking NodeJS agent is causing very high CPU utilization for NodeJS
applications being instrumented. To lower down the CPU consumption, I am
thinking
to less frequently send the data and send those without waiting for the
response using gRPC. So,
1. will try testing by increasing the maxBufferSize value from default 1000
to higher no.
2. Assuming Async gRPC IO is happening to send the metric/span/signal data,
is there a way I can ensure it is done in fire and forget manner; kind of
not having callback functions to process the response. I understand this
may not be reliable but we are ok with this.

Any input/guidance is highly appreciated.

Thanks -Pankaj