Right now benchmark results are showing a lot of variability, without making changes to apm-server.
The main suspect is elasticsearch running in the same machine as apm-server, as that demonstrated issues also in the past.
Regression tests have not being failing because of elastic/hey-apm#132 , but even with that fixed they can not be trusted.
Therefore we should move the elasticsearch instance that apm server uses to another machine.
it would be great to get this soon-ish, @elastic/observablt-robots feel free to reassign this back to me if you don't have the bandwidth!
Right now benchmark results are showing a lot of variability, without making changes to apm-server. The main suspect is elasticsearch running in the same machine as apm-server, as that demonstrated issues also in the past. Regression tests have not being failing because of elastic/hey-apm#132 , but even with that fixed they can not be trusted.
Therefore we should move the elasticsearch instance that apm server uses to another machine.
it would be great to get this soon-ish, @elastic/observablt-robots feel free to reassign this back to me if you don't have the bandwidth!