Keto Issues & Feature Requests

[KashifSaadat]

These are observations just from using the command line client. I've not gotten as far as logging into and validating the setup of the cluster internals yet.

Initial Boilerplate Setup:

• (Feature Request) Any benefit implementing the capability to setup a clean VPC for fresh deployments, so you can be up and running from scratch purely through Keto? Although we should be careful that functionality is not too closely coupled with a single cloud provider.

Notifications / Verbosity:

• (Feature Request) On Create / Delete cluster, list the resources to be modified
• (Feature Request) Add confirmation prompt on cluster modifications
• (Feature Request) Add '--force' flag or similar for non interactive mode (if the above is implemented)
• (Feature Request) On successful creation of each stack / resource group, write out to console (giving some level of progress on build) and a total of remaining resources to build.
• (Feature Request) Are there any specific resource groups that could be created in parallel to speed up deployment?

Labels / Naming:

• It appears that the additional labels specified weren't passed through in the build, or added onto the resources created. The following command was executed (keto --cloud aws create cluster kashtestcluster --ssh-key kash-key --networks subnet-33e01554,subnet-6988be31,subnet-e327d0aa --machine-type t2.medium --compute-pools 3 --pool-size 3 --labels kash1=tESt1,KASH2=teST2 --assets-dir ~/Downloads/certs/)
• (Feature Request) Ability to specify custom names per compute pool in the 'create cluster' command (not as crucial considering you can build compute and master pools independently)
• Primary network interfaces aren't 'Name' tagged

ELB:

• On build, all master nodes are out of service on the ELB. (Lewis mentioned this is a known issue on keto-k8 master)
• (Feature Request) Ability to specify source addresses for incoming connections, default to allow all.
• With a long cluster name, the address is truncated in creation of the ELB. I'm pretty sure this is a non-issue but wanted to raise just in-case it results in any weird edge case problems: https://keto-kashtestc-elb-z9cjw71ucl3v-9847412.eu-west-1.elb.amazonaws.com

Args / Errors:

• Clarify the expected key file names in the help output for '--assets-dir'
• "Error: unable to determine region" - (AWS Provider) Maybe provide additional supporting info to tell the user to make sure their AWS_PROFILE and AWS_DEFAULT_REGION env-vars are set.
• (Feature Request) Specify machine type independently for master nodes and different compute pools (not as crucial considering you can build compute and master pools independently)
• (Feature Request) In help output, include column with Y/N for required flags? You already mention default values in brackets, so that might be sufficient.
• Commands across create / delete don't seem consistent. Creating a master pool is specified as 'keto create masterpool NAME', however deletion is 'keto delete masterpool --cluster NAME'

Deletion:

• No validation checks on deleting a masterpool, when there are associated computepools
• ELB not deleted on deleting a masterpool
• Secondary data volumes not deleted on deleting a masterpool (10GB ones)
• Secondary network interfaces not deleted on deleting a masterpool
• Security Groups not deleted on deleting a masterpool

Creation:

• Can't create a masterpool without having a cluster created, can't create a masterpool in a cluster as one already exists. Is 'create masterpool' a valid command?

Questions:

• How does the coreos-version work? Does the string match against machine images located with the cloud provider? - Edit: Tested and yep it matches against the name.
• Master hosts have two volumes, the boot volume is increased when specifying '--disk-size' (which is correct according to the help command) and the additional volume is left at 10GB. What's the additional volume used for?