Healthcheck broken for streaming
Opened this issue · 0 comments
damienlaine commented
Description
When using SERVICE_MODE=websocket
, the current implementation does not properly handle health checks, leading healthcheck fail streak in orchestrated deployments.
Current Behavior
- Health checks fail when
SERVICE_MODE=websocket
is selected. The documentation is inconsistent:
Questions to Address
- Failback Implementation: Should we default to
SERVICE_MODE=http
and use it in combination withENABLE_STREAMING
to achieve streaming? - Potential Drawbacks:
- Could this approach introduce additional latency, system load, or other unexpected issues?
Expected Outcome
- A robust health check mechanism that works seamlessly, regardless of the
SERVICE_MODE
. - Clear and unified documentation of expected behavior for each
SERVICE_MODE
.