Restart ability of GraphScope
Open
enhancement
priority:low
- Dominant language
- C++
- Stars
- 3.6k
- Forks
- 468
- Avg merge
- 29m
- Merged PRs (30d)
- 1
Description
We launch a GraphScope cluster as follows:
1. Start the coordinator
2. The coordinator start the graph-manager, etcd, engine pods
3. The coordinator launch the GAE GRPC server
4. Ready for serve.
The problem is the GAE server may fail, and if it fails, the cluster will be shut down. Actually this is not necessary, as long as the coordinator pod is healthy, it can relaunch the other engine pods. This gives the ability to restart graphscope when failed, greatly reduced the vulnerability of the cluster.
Low-priority for now.
Contributor guide
Assessment
This issue has not been assessed yet.