metrics stop reporting

[deweyjose]

Hi,

We're currently seeing cases where after a week or so where metrics from some of our processes stop reaching Datadog. Restarting the processes will fix the issue, we see metrics land in Datadog immediately following a restart.

We've instrumented the client with an error callback, which is not getting triggered, we don't see any log entries for it. Has anyone else run into something like this or is unique to our experience?

This is how we're configuring the StatsD client.

We are on version 8.3.0.

NodeJS version 8.9.4.

new StatsD({
          cacheDns: true,
          prefix: 'someservice.',
          maxBufferSize: 10000,
          bufferFlushInterval: 10000,
          mock: false,
          errorHandler(error) {
            logger.error('Socket errors caught here: ', error);
          },
          globalTags: {
            env: 'prod'
          }
        });

[bdeitte]

Well that is pretty odd- I haven't heard of this one before. Do you have DD_AGENT_HOST set by any chance, and could that host IP change (and not be picked up because of cacheDns?). Not thinking of anything else from what you show here.

[hdavidzhu]

@deweyjose Dealing w/ something slightly similar -- if your maxBufferSize too big? Datadog agent's default buffer is 8KB, and what you have is 10_000 * 2B = 20KB

[abarre]

@hdavidzhu, sorry for asking a question on a thread from 2021. Why did you multiplie 10000 with 2B ?
maxBufferSize:8000 is not correct for datadog ?