The following is an example of migrating existing Lustre cluster in AWS to new subscription or new AMI. The same process can be used to upgrade Lustre to new version ('rolling upgrade').

First we will start with MGS:

From EC2 Dashboard, click on "Launch Configurations"


From "Launch Configurations" screen, click on checkbox with name that has "MGSLaunchConfig", then click on "Copy launch configuration"


From "Copy Launch Configuration from ...", click on "Edit AMI"


Once in the AMI select screen, search for AMI, in this case we use "loci" keyword, then click on "Select"


Select "Yes" on next screen to confirm the change


Click "Next: Configuration details" on next screen so we keep the type of EC2 the same. You could change it if you want.


Click "Skip to review" on next screen

Click "Create Launch configuration" on next screen


Click "Close" on next screen and go to EC2 Dashboard, click "Auto Scaling Groups"


On "Auto Scaling Groups" screen, select the checkbox with name that has "MGSNodes", then click Edit


On the next screen, change the "Launch Configuration" to the one just created (usually same name with 'copy' at the end if you did not change the name), then click "Save"


At the end of these steps, what we have done is added a new launch configuration to the auto scaling group. What we need to do next is to remove the old launch configuration


From "Launch Configurations" screen, click on checkbox with the old MGS Launch config (the one without 'copy' at the end), then from "Actions" menu, click "Delete launch configuration"


Confirm delete on next screen.


At this point we have the same MGS "Auto Scaling Groups" that only has the new Launch configuration with new AMI. To deploy the new configuration, we need to terminate the MGS instance so the Auto Scaling can launch new instance with new AMI.

NOTE: This will cause Lustre Filesystem to temporary paused (inaccessible) until the new instance is fully added to the cluster


From EC2 Dashboard, select "Instances", then click the checkbox with Name "mgs", then from "Actions" menu, click "Instance State"→ "Terminate"


Confirm "Yes, Terminate" on next screen

It should take approximately 2 - 5 minutes for a new instance to spin up.

EC2 Dashboard will have the old mgs instance terminated, and new mgs instance running. Lustre Filesystem should be operational.


Above is example of how to upgrade/migrate to new Lustre or AMI on MGS only. The same procedure applies to MDS and OSS (Preferably one server at a time)

  • No labels