I’ve installed a vanilla version of the latest Helm chart for Confluent Platform in both Docker for Windows and in AKS. It installs CP 6.0.1 by default. In both cases, the vanilla installation had about 36k/s message traffic before I even added my own producers. Is this normal? What is generating the traffic? Is there a way to tune it?
For kicks, I decided to tweak the Helm Chart to install CP 6.1.1 instead to see if the newer version fixed it, but the traffic actually doubled, almost to the byte. I see that there are now 4 consumers listed in the Consumer manager, all Control Center, two with “6-0-1” in their name and two with “6-1-1” in their name. Did my naïve upgrade via Helm fail to remove part of the 6.0.1 installation?
I’ve tried to google this a little but either there isn’t much out there on the topic or I can’t narrow my search terms sufficiently. Any help is appreciated!
Control Center does have some traffic associated with the internal streams application. For comparison, when starting Confluent Platform with a local tarball install, e.g. “confluent local services start”, it also shows a baseline traffic ~20Kbps, so sounds relatively normal.
As to why there is both “6-0-1” and “6-1-1”, maybe this is related to the doubling in traffic? Maybe start with a clean environment and spin up just “6-1-1”?
I did as you suggested, reinstalling 6.1.1 from scratch. Now Control Center is failing with this error:
"GET /2.0/metrics/TwpZ-iMcQASyuZP3gDykKA/broker/detail HTTP/1.1" 403 62 5 (io.confluent.rest-utils.requests)
ERROR Request Failed with exception (io.confluent.rest.exceptions.DebuggableExceptionMapper)
javax.ws.rs.ForbiddenException: must have broker metrics access
and a big stack trace. I’ve seen this error before too and also can’t find anything on the web. Last time a clean reinstall fixed this one, but that didn’t work this time (I tried this twice). Any idea what this problem is?
Up to six retries and I’m still getting this error. Thinking I will just disable Control Center for now since I don’t have a lot of time to look into this. I filed an issue on the CP helm chart project. Thanks for your help.