[
https://issues.apache.org/jira/browse/CLOUDSTACK-7841?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14197241#comment-14197241
]
Sheng Yang commented on CLOUDSTACK-7841:
----------------------------------------
Verification steps:
Positive verification on LB rules:
1. Start two vms in a network. Open port 22 in firewall. Set LB rules for 22
for two vms.
2. Ssh using public IP, then doing ping gateway or other commands which able to
continuous monitor the connectivity.
3. Add a new LB rule in the same network.
Before fix: the existing connection would be dropped.
After fix: the existing connection would always works.
Negative verification on LB rules after fix:
1,2. Same as step 1,2 above.
3. Log into VR, run: " netcat <public_ip> -l -p 10000 ", which would listen on
port 10000, and would result in haproxy fail to listen on the port later.
4. Assign LB rule on port 10000. It would failed to apply since haproxy won't
able to listen on port 10000. But the existing connection still works after fix.
> Existed connections are disconnected when update load balancer configuration
> ----------------------------------------------------------------------------
>
> Key: CLOUDSTACK-7841
> URL: https://issues.apache.org/jira/browse/CLOUDSTACK-7841
> Project: CloudStack
> Issue Type: Bug
> Security Level: Public(Anyone can view this level - this is the
> default.)
> Components: Network Devices
> Reporter: Sheng Yang
> Assignee: Sheng Yang
> Fix For: 4.5.0
>
>
> Applying load balancer rules breaks existing connections and causes short
> outage.
> That's because our currently logic of handling haproxy reload configuration
> is not graceful enough, and focused on how to recover from failed newly
> configuration.
> There would be a way to improve this.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
