search

EC-Release / sdk

The Agent SDK
Other
4 stars 7 forks source link

Bug: Deploy Server Agents Across VMs #110

Closed ayasuda2OO3 closed 4 years ago

ayasuda2OO3 commented 4 years ago

Use Case

Current key-value format for the agent superconn establishment as follows-

agentId+refId+groupId

It's very possible a superconn be replaced by one another from a different server Agent, which may revoke the server-agent HA, and cause the instability in the connectivity.

Proposal

1) Find the root cause of the duplicate or co-existed superconns in the current behaviour. (v1.1,212) 2) differentiate the superconns across different VMs. 3) Make server HA, as is in the CF environement.

ayasuda2OO3 commented 4 years ago

Findings

1) In the prod release (v1.212) the refId used in the superconn which is the process#, hence no duplicate issue. 2) as for superconns, the process# is invisible in the health endpoint in the current prod release. 3) the refId assignment of the server inst is the intance# in the coming watcher release. watcher is designed as the replacement for the server agent for container scaling.

Close the issue for the clarity. @palokam