Conversation
googlebot
added
the
cla: yes
| // Istio service mesh. | ||
| message MeshConfig { | ||
| // Address of the egress Envoy service (e.g. _istio-egress:80_). | ||
| string egress_proxy_address = 1; |
There was a problem hiding this comment.
why is egress_proxy_address being removed as part of the mixer_address separation?
There was a problem hiding this comment.
Ah shux.. It was an unused config item.. We got rid of egress proxy a long time ago. Thought I would remove it as well.
| // reported to the mixer for HTTP requests and TCP connections. Default | ||
| // Address of the server that will be used by the proxies for policy | ||
| // check calls (e.g. _istio-mixer:15004_). By using different names for | ||
| // policyCheckServer and telemetryServer, it is possible to have one set |
There was a problem hiding this comment.
Can we use mixer_check_server and mixer_report_server for consistency with the mixer API service?
There was a problem hiding this comment.
That is a good point. What I am trying to do here is to keep the config at concept level (policies, telemetry, auth, etc.). There are other parts of the config that are essentially pointing to implementation specific things, which to a layman is going to be hard to understand. If I am starting off with Istio, I look at global config and turn off things I don't want/enable things I want. Make sense?
(I am waiting to get rid of those RDSrefresh delays, etc. but haven't found a good way to do it yet).
There was a problem hiding this comment.
Somewhat. We seem to be conflating mesh administrator configuration and operator policy config into the same protobuf which makes this difficult. Configuring whether a feature (e.g. auth, telemetry) is enabled globally (or per namespace / service) is a different concept than how the services that enable those features are accessed by the proxies.
|
Lets not merge this until all tests pass in istio/istio for my PR. Dont want to upset stuff.. |
* separate check and report clusters * fix * nits * nit * backward compat
5 participants
Pursuant to PR #358 ..
TODO: Once this is implemented in Pilot such that policy_check_cluster and telemetry_cluster are always specified, we need to