javier-ruiz-voxsmart commented on issue #12747:
URL: https://github.com/apache/apisix/issues/12747#issuecomment-5003171948
## Resolution — Full Root Cause and Fix (Kubernetes + Helm + Ingress
Controller)
After extensive debugging, I'm sharing the full resolution for anyone
hitting this in a Kubernetes environment with the APISIX Helm chart and
ingress-controller.
### Root Cause
The Datadog plugin reads `host` and `port` from **`plugin_metadata`**
(stored in etcd), **not** from `plugin_attr` in `config.yaml`. This is the key
distinction @Baoyuantop pointed out.
However, in a Kubernetes + Helm + ingress-controller setup, there are
additional gotchas:
1. **`pluginAttrs.datadog` in Helm `values.yaml` is useless** — it renders
to `plugin_attr` in `config.yaml`, which the Lua plugin ignores at runtime. The
plugin defaults to `127.0.0.1:8125`.
2. **Setting `plugin_metadata` via the Admin API (`PUT
/apisix/admin/plugin_metadata/datadog`) gets wiped** — the ingress-controller
reconciliation loop overwrites it.
3. **Setting `host`/`port` in `ApisixGlobalRule` config does nothing** —
`host` and `port` are metadata-level settings, not route/global config-level.
### The Fix — `GatewayProxy.spec.pluginMetadata`
The correct declarative approach in Kubernetes is using the `GatewayProxy`
CRD (`v1alpha1`), which has a `pluginMetadata` field that the
ingress-controller syncs to etcd automatically:
```yaml
# In Helm values.yaml under ingress-controller:
ingress-controller:
gatewayProxy:
createDefault: true
pluginMetadata:
datadog:
host: "datadog.monitoring.svc.cluster.local" # Your DogStatsD
service
port: 8125
namespace: "apisix"
constant_tags:
- "source:apisix"
- "service:apisix"
- "team:myteam"
```
Additionally, enable the plugin globally via `ApisixGlobalRule`:
```yaml
apiVersion: apisix.apache.org/v2
kind: ApisixGlobalRule
metadata:
name: global-default
spec:
ingressClassName: apisix
plugins:
- name: datadog
enable: true
config:
prefer_name: true
```
### Summary
| What | Works? | Why |
|------|--------|-----|
| `pluginAttrs.datadog` in Helm values | No | Renders to `plugin_attr` in
config.yaml — ignored by Lua plugin |
| Admin API `PUT /plugin_metadata/datadog` | Temporarily |
Ingress-controller reconciliation overwrites it |
| `ApisixGlobalRule` config `host`/`port` | No | `host`/`port` are metadata,
not config |
| **`GatewayProxy.spec.pluginMetadata`** | **Yes** | Synced to etcd
`plugin_metadata` by ingress-controller — survives reconciliation |
### Versions Tested
- APISIX Helm chart: 2.16.0 (APISIX 3.17.0)
- Ingress Controller: 1.2.0
- Datadog Agent: 7.x (chart 3.230.0)
Hope this helps anyone running into the same `connection refused` errors
with the Datadog plugin in Kubernetes.
--
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]
