bitflicker64 commented on issue #3137:
URL: https://github.com/apache/hugegraph/issues/3137#issuecomment-5170803498
Possible implementation roadmap, splitting the proposal above into
independently reviewable steps:
```text
Current problem
|
v
Graph creation is eventually consistent across Server replicas.
CreateGraph returns before the graph is usable cluster-wide.
```
**PR 1 (small, low risk)**
- Make the creating Server wait for its own GRAPH_CREATE event: replace
`notify()` with `notifyAndWaitEvent()` on the PD path, matching what the local
path already does.
- Guarantees the creating Server can serve the graph the moment CreateGraph
returns.
**PR 2 (medium)**
- Introduce graph readiness reporting: each Server reports READY / LOADING /
FAILED for a graph to PD.
- Expose cluster-wide creation status through an API that clients can await,
closing the cross-replica window.
**PR 3 (larger, architectural)**
- Move graph creation ownership entirely into PD; Servers become pure
replicas converging on PD metadata.
- The special "creating Server" path disappears, and the failure modes in
the risks section above (log-only convergence failure, config persisted before
schema preparation) go away by construction.
PR 1 is independently useful and could land while the rest is discussed. I
am happy to start on it if the direction sounds right.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
