[
https://issues.apache.org/jira/browse/CLOUDSTACK-4856?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13826700#comment-13826700
]
Sheng Yang commented on CLOUDSTACK-4856:
----------------------------------------
A pair of RvR would be distributed on different hosts, and only update one host
at time would likely cause chaos in consistency of RvR status.
E.g. The switch over happen, and the first host contained old BACKUP which
already switch to the MASTER. Then the update status of the host would return
MASTER.
At this time, we haven't updated host of the old MASTER yet, so we don't know
if the states of RvR are both up to date. And we would see two MASTER in the
network as the result.
That's the reason why we always update one pair of RvR(not individually one of
them), which we can know after update complete, we would know the latest status
of RvR in the network. This make things more simple.
> Optimize on the # of control commands sent by MS to HV host
> -----------------------------------------------------------
>
> Key: CLOUDSTACK-4856
> URL: https://issues.apache.org/jira/browse/CLOUDSTACK-4856
> Project: CloudStack
> Issue Type: Bug
> Security Level: Public(Anyone can view this level - this is the
> default.)
> Components: Management Server
> Affects Versions: pre-4.0.0, 4.1.0, 4.2.0
> Reporter: Koushik Das
> Fix For: 4.3.0
>
>
> This is a placeholder bug for all control commands whose count can be
> minimized by doing some optimisation.
> 1. CheckRouterCommand
> This is used for checking status of redundant VR and is send for each router
> at regular intervals. Instead of sending individual commands to each VR,
> these can be batched for all routers residing on the same host.
> This is a general idea and can be extended to other such commands as well.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
