Test neutron of openstack (JUNO), openstackneutron

Test neutron of openstack (JUNO), openstackneutron

I tested the company's openstack environment over the past two days because the environment was built by other colleagues. I am only responsible for the relevant tests, and one server is installed independently.

Controller node, the other two servers are installed with network node + compute node. The general environment is as follows:

The Internet IP address segment is 192.168.1.0/24, of which 192.168.1.100 ~ 192.168.1.200 is the allocation range of floatingip, and the Intranet IP segment is 172.16.1.0/24.

On compute0, there is an instance VM00 with an Internet IP address 192.168.1.102 and an Intranet IP address of 172.16.1.2. Compute1 has an instance

VM01, assigned the Internet IP address 192.168.1.103, and the Intranet IP address is 172.16.1.3. VM00 and VM01 are on the same Intranet subnet.

The Intranet and Internet are interconnected through the router, so you can communicate with VM00 and VM01 on the controller.

Now you need to remove compute0, You need to migrate the VM00 to compute1, and then disable compute0. Use nova migrate and nova

After the resize-confirm command migrates VM00 to compute1 and shutdown compute0, it cannot be synchronized with VM00 and VM01 on the controller.

Line communication.

The reason is as follows:

Before migrating a virtual machine, the internal network composition of the two compute instances is as follows:

There is a namespace of the router and dhcp server on compute0. If you migrate the VM00 to VM01, shutdown compute0. From the point of view, there is no router namespace,

Therefore, the routes to VM00 and VM01 are different. The following figure shows the network structure after migration:

Why does the dhcp server still exist? This may be related to the openstack installation. It may be that the dhcp server is in HA mode during the installation, so you will certainly ask, why do I know that the dhcp server exists?

? I create an instance and use dhcp mode. The new instance is assigned to an ip address, and the ip netns shows that the dhcp namespace is generated.

Then how can I disable compute0 to make the network still valid? Naturally, I want to migrate the router. The solution is similar. I use the HA router and know to use the -- HA True option when creating the router.

After the HA router, the two compute networks are composed of the following:

After the HA router is created, a namespace of the router is created on compute1. After the VM00 is migrated and compute0 is shut down, the following figure shows the network structure of the two compute nodes:

The router of compute1 takes over for several seconds after compute0 is disabled to Ensure network connectivity. If a VM is created on compute1, the name namespace of the dhcp server will also

It indicates that the new dhcp server can continue to work.

(Because the company cannot send a blog post to external users and can't post the operation content, it can only be recorded in its mind. Come back and summarize and write this blog post)