We’ve been upgrading from Kong 0.11 to Kong 2.2.1 and we started to see an extremely high upstream latency for our client-facing service.
The rest of the services (server-to-server) are working well.
The metrics are exported to Datadog via their plugin and the metrics name that we use is kong.upstream_latency (95percentile, avg and max).
Currently we rolled back to the old Kong version (0.11) and the upstream latency is fine.
Do you have any ideas where can I start from ?