AlinsRan opened a new pull request, #13991:
URL: https://github.com/apache/apisix/pull/13991
### Description
In multi-cluster mode, Kubernetes service discovery can only resolve a
`service_name` against one cluster (`id/namespace/name:port_name`). This PR
lets an upstream pick several clusters explicitly and get the union of their
endpoints:
```json
{
"discovery_type": "kubernetes",
"service_name": "default/plat-dev:port",
"discovery_args": {
"cluster_ids": ["release", "staging"]
}
}
```
Behaviour:
- The nodes are the union of the matching endpoints in the listed clusters
(an `id` from the `discovery.kubernetes` array), cached per cluster endpoint
version, so endpoint changes in those clusters keep refreshing the upstream.
The same `host:port` found in more than one cluster is used once.
- Clusters that are not listed, including clusters added to the
configuration later, never contribute nodes.
- If none of the listed clusters has matching endpoints, the upstream goes
through the existing "no valid upstream node" path. There is no fallback to
other clusters.
- An unknown cluster id is logged as an error and skipped at runtime. The
Admin API has no hook that validates `discovery_args` against the loaded
discovery configuration (`check_upstream_conf` only runs the schema and
field-level checks, and the discovery modules only expose `nodes` /
`init_worker` / `dump_data`), so it is not checked when the configuration is
submitted.
- With `cluster_ids`, `service_name` must not carry a cluster id prefix. The
Admin API rejects `cluster_id/namespace/name:port_name` in that case, and the
discovery module logs an error and returns no nodes if such a configuration
arrives another way.
- In single-cluster mode, `cluster_ids` has no cluster to match, so the
upstream gets no nodes and an error is logged.
- Without `cluster_ids`, the existing behaviour is unchanged
(`namespace/name:port_name` in single-cluster mode,
`id/namespace/name:port_name` in multi-cluster mode).
`cluster_ids` is added to `discovery_args` in the upstream schema next to
the existing `namespace_id` / `group_name`, as a non-empty array of unique
strings.
Tests: `t/kubernetes/discovery/kubernetes5.t` covers selection against the
kind cluster (listed vs unlisted clusters, unknown ids, endpoint updates), the
union and dedup with fixed endpoint data, a proxied request through a route
with `cluster_ids`, single-cluster mode, and the Admin API validation.
#### Which issue(s) this PR fixes:
Fixes #
### Checklist
- [x] I have explained the need for this PR and the problem it solves
- [x] I have explained the changes or the new features added to this PR
- [x] I have added tests corresponding to this change
- [x] I have updated the documentation to reflect this change
- [x] I have verified that this change is backward compatible (If not,
please discuss on the [APISIX mailing
list](https://github.com/apache/apisix/tree/master#community) first)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
