If we poll the master's /state endpoint on large clusters, we can
degrade master performance. Since the /state endpoint is used to
generate per-agent resource metrics, and we can already get those
by running the exporter on each agent, add a new flag that can
disable the collection of these metrics.
@gkleiman @philipnrmn This PR adds a flag with the same functionality as in #69. I'd prefer to make this change independently of the other changes in the #69 branch.
If we poll the master's
/state
endpoint on large clusters, we can degrade master performance. Since the/state
endpoint is used to generate per-agent resource metrics, and we can already get those by running the exporter on each agent, add a new flag that can disable the collection of these metrics.This is equivalent to the fix in #69