Hello again,
Recently my server running openstack-compute had to be brought down for maintenance (which was running some 60 VMs). This is when i learnt how to use "nova evacuate"
This is a cool feature in openstack using which you can get back instances from the server which has gone down. Ya, but it requires some prior configuration.
The important stuff is how you keep you root disk. Openstack allows three different configurations to fetch the root disk.
1. Root on volume (EBS volumes) - By this instance can be brought up in other server with the same EBS(root) volume.
2. Instances folder in shared mode using NFS, GlusterFS, etc - Basically the disk file needs to be accessed across all nodes, i.e mount point for instances folder must be shared across all nodes. GlusterFS appears to work super cool in this method.
3. Disks are cleaned but instances are recreated from fresh disk from glance - Even though the disk is ultra new, instance information (like IP, instance Id,etc.,) are retained just as it was before evacuation.
Once the server is down, all you need to do is issue the following command
# nova host-evacuate --target_host <server where the VMs need to be transferred> --on-shared-storage <server which got down>
Eg:
# nova host-evacuate --target_host old_node --on-shared-storage new_node
I have added a video demonstration of the entire flow. I hope this helps someone. Have a great day
:) :) :) :) :) :) :) :)
Recently my server running openstack-compute had to be brought down for maintenance (which was running some 60 VMs). This is when i learnt how to use "nova evacuate"
This is a cool feature in openstack using which you can get back instances from the server which has gone down. Ya, but it requires some prior configuration.
The important stuff is how you keep you root disk. Openstack allows three different configurations to fetch the root disk.
1. Root on volume (EBS volumes) - By this instance can be brought up in other server with the same EBS(root) volume.
2. Instances folder in shared mode using NFS, GlusterFS, etc - Basically the disk file needs to be accessed across all nodes, i.e mount point for instances folder must be shared across all nodes. GlusterFS appears to work super cool in this method.
3. Disks are cleaned but instances are recreated from fresh disk from glance - Even though the disk is ultra new, instance information (like IP, instance Id,etc.,) are retained just as it was before evacuation.
Once the server is down, all you need to do is issue the following command
# nova host-evacuate --target_host <server where the VMs need to be transferred> --on-shared-storage <server which got down>
Eg:
# nova host-evacuate --target_host old_node --on-shared-storage new_node
I have added a video demonstration of the entire flow. I hope this helps someone. Have a great day
:) :) :) :) :) :) :) :)
No comments:
Post a Comment