#2220 clm: Main thread of node agent is blocked due to synchronous message sending

[Anders Widell]

The CLM node agent performs a synchronous send with a long time-out towards the CLM server. If there is no response from the CLM server, this could cause the main thread in the CLM node agent to be blocked for a long time.

[Mathi Naickan]

What is the problem because of this blocking send? Till election was introduced, this blocking send is by design.
Have you come across any new issues in the context of headless scenarios?

[Anders Widell]

If no reply is received (e.g. because of message loss or because the CLM server died before being able to reply), then the CLM node agent is stuck in the blocking send until it times out. No election can happen on this node, and it cannot send a new join request either in case a new CLM server appears.

[Anders Widell]

changeset: 8452:15662fec6dc9
branch: opensaf-5.0.x
parent: 8447:c0ccb717d4e7
user: Anders Widell anders.widell@ericsson.com
date: Mon Dec 19 14:40:51 2016 +0100
summary: clm: Send MDS messages asynchronously to avoid blocking the main thread [#2220]

changeset: 8453:73218db7384f
branch: opensaf-5.1.x
parent: 8450:e9f73afbfea3
user: Anders Widell anders.widell@ericsson.com
date: Mon Dec 19 14:45:46 2016 +0100
summary: clm: Send MDS messages asynchronously to avoid blocking the main thread [#2220]

changeset: 8454:a7be9cd38ab7
parent: 8451:8196ad473f78
user: Anders Widell anders.widell@ericsson.com
date: Mon Dec 19 14:45:46 2016 +0100
summary: clm: Send MDS messages asynchronously to avoid blocking the main thread [#2220]

[staging:15662f]
[staging:73218d]
[staging:a7be9c]