use the Prometheus /-/healthy
endpoint for the default value for health_check_url
#6966
Labels
backlog
Triaged Issue added to backlog
backport completed
Issue PRs have been backported
backport needed
Issue PRs require backport to versions specified in comments
enhancement
This is the preferred way to describe new end-to-end features.
by default, use the
/-/healthy
endpoint that Prometheus has, rather than the main prometheus root url.This is because sometimes Prometheus doesn't have the UI installed and this causes the root URL to tell the client to redirect to an invalid
/graph
endpoint, causing a error code to be returned, thus Kiali UI will report Prometheus being unreachable which isn't true.For example, in one setup we've seen in testing, this happens.
If I go into the prometheus pod itself, and try to submit a request to check its health (using curl), I get this:
Which looks fine, but that is actually returning a 302 - it is redirecting the client to
/graph
and if I go to the/graph
path I see this:and that returns a error code of 500, which results in the Kiali UI to report Prometheus as "unreachable". But that is not true. Kiali UI can show the graph and metrics fine. This is because the Prometheus backend is working fine, it is just the Prometheus UI isn't installed or configured properly. But Kiali doesn't care about that - it just cares that Prometheus can respond to our queries (which it can in this particular case).
The text was updated successfully, but these errors were encountered: