Skip to content

Homework 1. Multi Node

life1347 edited this page Oct 24, 2013 · 6 revisions

OpenStack Version : Grizzly

OS Version : Ubuntu 13.04 server 64bit 3.8 kernel

OVS Version : 1.9

Basically, I followed the mseknibilel tutorial to install multi-node OpenStack.

ovs-datapath rebuild

Before I choose ubuntu 13.04 as my host OS, I used 12.04 with 3.8 kernel. But, I face the problem that OpenvSwitch could not establish tunnel. After search such problem on google, I found the last paragraph of this page saying that dkms for 12.04 was still under testing.

So we have two ways to solve this problem:

  • still use 12.04 but with 3.5 kernel.
  • update to 13.04

Here, I choose later one to solve it. But we still need to rebuild ovs-datapath with under commands

$ sudo apt-get install -y openvswitch-datapath-source module-assistant
module-assistant prepare
$ sudo cd /lib/modules/`uname -r`/build/include/linux
$ sudo ln -s ../generated/uapi/linux/version.h .
$ sudo module-assistant auto-install openvswitch-datapath

Volumes set up

Follow steps on tutorial, add script to /etc/rc.local in order to mount volumes after server reboot.

$ sudo losetup /dev/loop2 %Your_path_to_cinder_volumes%
//%Your_path_to_cinder_volumes% is the path that cinder_volumes store.

Current Problems

dhcp failed The problem now is that vm cannot find dhcp after launching. After testing find out that network node could not send arp reply to compute, however, sometimes after reboot compute node and network node, suddenly everything work fine, so I suspect is that ovs gre tunnel not work properly, but I still tried to solve this problem.

SOLUTION:

With TA's help, and rebuild networks likes DevStack default networks setting, there are two places we have to notice:

  1. Make sure that every nodes' time is correct. In this case, network node's dhcp reply packet could not send to compute node and it was caused by time difference between two nodes.
$ sudo /etc/init.d/ntp stop
$ sudo ntpdate time.stdtime.gov.tw
  1. Sometimes dhcp interface is not active, we have to restart dhcp agent manually.
$ sudo /etc/init.d/quantum-dhcp-agent restart

OpenStack Snapshot

ops1 ops2 ops3 ops4

Clone this wiki locally