I have seen a few users that have attempted to deploy multiple Log Insight VMs, but did so by cloning a bootstrapped Log Insight instance. In this post, I will talk about bootstrapping and why such instances should not be cloned.
By default, a Log Insight virtual appliance comes in bootstrapped mode. When you actually first power on a new Log Insight instance it goes through a first boot process — first boot is separate from bootstrapped mode. This first boot process does a variety of things including configure the /storage/core and /storage/var partitions as well as install Log Insight. The process takes about 5 minutes + the time to format the storage (assuming 2 TB the format will take about 20 minutes) and results in a system reboot near the completion of the operation.
Once the first boot — which as the name applies only happens on the first boot — completes, the system is left in a bootstrapped mode. You can confirm bootstrap mode by navigating to the UI of the LI instance. If you are presented with a Welcome screen that then asks whether you wish to deploy a new instance or join an existing instance then you know you are bootstrap mode, if you are asked to provide the admin credentials or you are asked to authenticate then you know you are not in bootstrap mode.
Cloning a Bootstrapped Instance
The reason why you do not want to clone a bootstrapped instance is because the instance will always be standalone — you will not be able to create a cluster with the VM template. In addition, certain unique identifiers will be created which may cause issues later down the road. Now if you wish to just clone and deploy standalone instance then perhaps this model will work for you, but I would say for most this is undesirable. How do you fix the template? The only supported way to get to bootstrapped mode is to deploy a new instance and ensure you do not complete the bootstrap process — put another way, once you bootstrap a node there is no supported way to go back.
Creating a LI Template
To create a LI template you have two options. First you could just take the base OVA and clone it out — the problem with this approach is that you will need to go through the firstboot process — again separate from the bootstrapped mode — for every clone which makes the operation take longer than it needs to. The alternative and best practice would be deploy the OVA, power it on and go through the first boot process. Once the LI splash screen is shown, power down the VM and you have your template. Long story short, just do not go to the UI and configure the node prior to creating the template.
Note: This information is also important for those of you deploying LI to vCloud Director. For more information, see this post.
© 2015, Steve Flanders. All rights reserved.