Controllers ipv4 forwarding disabled when neutron and rabbitmq are decomposed
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
In Progress
|
High
|
Michael Polenchuk |
Bug Description
Fuel 9.2 - Control plane split into controllers, neutron nodes, and rabbitmq nodes.
Seems that all tasks which are setting ipv4 forwarding were delegated into decomposed nodes.
Controller left with:
root@node-107:~# cat /proc/sys/
0
When network templates being used, it causes a lot of troubles as namespaces are not working properly, thus almost all services are not able to communicate.
In my lab deployment (all nodes virtualized, no network templates, simple network) ip forwarding is disabled too, although everything is working fine.
Task graph should be reviewed and verified which task should be applied on controller to enable forwarding when neutron and rabbit are decomposed.
Changed in fuel: | |
assignee: | nobody → Michael Polenchuk (mpolenchuk) |
milestone: | none → 12.0 |
tags: | added: area-library |
Changed in fuel: | |
status: | New → Confirmed |
importance: | Undecided → High |
Workaround:
Add to deployment_ tasks.yaml of any plugin: setup_repositor ies] ip_forward= 1; sed -i 's/#net. ipv4.ip_ forward= 1/net.ipv4. ip_forward= 1'
- id: fix_forwarding
type: shell
role: [controller, primary-controller, rabbitmq, primary-rabbitmq]
requires: [plugins_
required_for: [sync_time]
parameters:
cmd: sysctl -w net.ipv4.
retries: 3
interval: 20
timeout: 180