Revise FEATURES.rst to reflect the reality

We have spent a lot energy convincing Octavia team to 1) fix node status
update bug; 2) emit event notifications when node status change. Neither
of these work was smooth due to Octavia team's ignorance. We have to
hold back these features hoping that they will change their minds
someday.

Change-Id: I1a4c28fe99b354ed5c1453ee0a57343d588e9e9d
This commit is contained in:
tengqm 2017-04-17 21:34:53 -04:00
parent d23a08d143
commit 41d75e554d

View File

@ -144,6 +144,24 @@ message which carried its handling capacity and said it want to assume
some health registries.
Host Fencing Support
--------------------
To ensure a seemingly dead node is actually dead, all HA solutions need a way
to kill a node for sure. Senlin is no exception here. We have support to force
delete a VM instance already. The need is a mechanism to kill a failed host.
LB HealthMonitor based failure detection
----------------------------------------
Ideally, Senlin could rely on the LBaaS service for node failure detection
rather than reinventing the wheel. However, LBaaS (Octavia) is not fixing the
obvious bug.
Another option is to have LBaaS emit events when node failures are detected.
This proposal has failed find its way into the upstream.
When the upstream project (Octavia) has such features, we can enable them from
Senlin side.
Low Priority
~~~~~~~~~~~~