discovery fixed policy breaks if service stays down for extended period
Affects | Status | Importance | Assigned to | Milestone | ||
---|---|---|---|---|---|---|
Juniper Openstack | Status tracked in Trunk | |||||
Trunk |
In Progress
|
Undecided
|
Deepinder Setia |
Bug Description
Discovery has fixed policy for service assignment in which services are assigned to consumers in a fixed, static or constant manner. For example if there are "n" publishers of a service and there are "m" consumers that are interested in "k" instances (say 2) of service, then all "m" consumers will get <n0, n1, n2 ... nk> service instances. This is akin to priority order. If an instance, say "ni" such that 0 <= i <= k went down for an extended period (> 15 seconds) and comes back up, it should no longer be assigned to a new consumer because it should go to the bottom of the prioritized list. It should not retain its position. Current behavior of discovery does not match this expectation and "ni" service continues to get assigned to new consumers.
affects: | opencontrail → juniperopenstack |
Reviewed: https:/ /review. opencontrail. org/5514 github. org/Juniper/ contrail- controller/ commit/ 323c3ba37bd1ff2 d68a0da11ccee7c 6dc39d189c
Committed: http://
Submitter: Zuul
Branch: master
commit 323c3ba37bd1ff2 d68a0da11ccee7c 6dc39d189c
Author: Deepinder Setia <email address hidden>
Date: Wed Dec 10 15:09:27 2014 -0800
Bug 1401304: Reset sequence number when service is republished. This should push service instance to bottom of priority list
Change-Id: I499334d55db44e b274bb7e1f7b69d fd38ca83ec3