LKEBot - A Slack Bot for Deploying Lab LKE Clusters

This is a Slack Bot for managing a dynamic fleet of Linode LKE clusters, with an upper limit on deployment counts and sizing, geared towards chat-ops teams who want short-lived, ephemeral Kubernetes clusters for testing and deployment.

Running the Bot

To run the bot manually, first build the software:

go build .

and then execute the resulting lkebot binary, with the appropriate flags:

./lkebot --slack $SLACK_TOKEN --linode $LINODE_TOKEN

To run this in Docker, you can use the filefrog/lkebot image, and your foavorite orchestrator / command-line flags:

docker run --rm -d \
  -e LKEBOT_LINODE_TOKEN=$LINODE_TOKEN \
  -e LKEBOT_SLACK_TOKEN=$SLACK_TOKEN \
  filefrog/lkebot:latest

For Kubernetes deployments, we have a sample YAML file, in k8s/deploy.yml; modify that how you see fit (namespace and configuration), and then apply it:

kubectl apply -f k8s/deploy.yml

Interaction

In direct one-on-one messages, commands can be sent as-is:

If you invite the bot to a channel, channel members will be able to interact with it by prefixing the command with a lone @-name:

The following commands are known:

• help - Print a helpful message about known commands.
• info - Print the chatbots configuration.
• list - Print the list of currently deployed clusters.
• deploy ... - Deploy a new cluster.
• renew ... - Renew the lease on a cluster.
• expire ... - Immediately release the lease on a cluster.
• teardown ... - Immediately delete a cluster.
• access ... - Retrieve the $KUBECONFIG for a cluster.

help

Prints out a helpful reminder of how to interact with the chatbot.

This command takes no arguments.

info

Prints out the chatbot configuration, including defaults and upper limits.

This command takes no arguments.

list

Queries the Linode API (caching the results) and then prints out the current status of deployed clusters, their sizing, and expiration times (relatively speaking).

This command takes no arguments.

deploy

Deploys a new LKE Kubernetes cluster.

Usage:

deploy NAME [-v K8S-VERSION] [-r REGION] [-i INSTANCE] [-n NODES] [--for HOURS]

The following flags are defined:

• -v or --version - The version of Kubernetes to deploy.

• -r or --region - The name of the Linode region to deploy the LKE cluster into. This region must support "Kubernetes".

• -i or --using - The name of the Linode instance type to deploy in this cluster's node pool.

• -n or --nodes - How many worker nodes (pursuant to operator-imposed upper limits) to deploy.

• --for - How long (in hours) should this cluster live.

If any or all of these options are omitted, suitable built-in and operator-provided defaults are used.

renew

Renew the lease on a deployed cluster.

Usage:

renew NAME [--for HOURS]

The following flags are defined:

• --for - How many more hours to add onto the clusters lease. If not specified, defaults to 1 (hour).

expire

Immediately expires a cluster's lease, making it sweepable (without explicitly tearing it down).

Usage:

expire NAME

This command takes no additional arguments.

teardown

Immediately tear down a cluster, deleting its worker nodes, and obliterating all workloads running on it.

Usage:

teardown NAME

This command takes no additional arugments

access

Retrieves and prints a suitable KUBECONFIG for accessing this cluster via the kubectl CLI.

Usage:

access NAME

This command takes no additional arugments

Configuration

You can configure the bot via either command-line flags or environment variables. Here's the list of knobs and levers you can fiddle with:

• $LKEBOT_SLACK_TOKEN, or --slack - REQUIRED - Sets the Slack API token that the chatbot will use to chat.

• $LKEBOT_LINODE_TOKEN or --linode - REQUIRED - Sets the Linode API token that the chatbot will use to deploy and tear down LKE clusters.

• $LKEBOT_SWEEP_INTERVAL or --sweep - Sets the interval, in minutes, that the bot should look for expired clusters, and schedule their decommission. Defaults to 1; setting it to 0 effectively disables sweep logic.

• $LKEBOT_CHANNEL or --channel - Sets the name of the channel to report expired cluster teardown to. Has no effect if sweeping is disabled. If you set this to the empty string (which also happens to be the default), sweep messages will be printed to stdout, instead of to Slack.

• $LKEBOT_MAX_CLUSTERS or --max-clusters - An upper limit on how many LKE clusters this chatbot is allowed to deploy. Defaults to 5.

• $LKEBOT_MAX_NODES or --max-nodes - An upper limit on how big of an LKE cluster this chatbot is allowed to deploy. Defaults to 5.

• $LKEBOT_DEFAULT_REGION or --default-region - What Linode region should clusters be deployed to if no explicit region is set? Defaults to us-east; make sure the region you choose has "Kubernetes" available.

• $LKEBOT_DEFAULT_INSTANCE or --default-instance - What Linode instance type to use in the node pool for new clusters. Defaults to g6-standard-2; make sure the instance type you choose actually exists.

• $LKEBOOT_DEFAULT_SIZE or --default-size - How large of a cluster should the chatbot deploy, if no size is explicitly set? Must not execeed the set upper limit for cluster size.

• $LKEBOT_DEFAULT_K8S_VERSION or --default-k8s-version - What version of Kubernetes should the chatbot deploy, when not told expressly which to use? Defaults to 1.18; make sure the version you specify is known to Linode / LKE.

• $LKEBOT_BLACKLIST_CLUSTERS or --blacklist - One (or more) cluster names to explicitly prohibit the chatbot from touching. The command-line flag can be set multiple times; the environment variable is a comma- or space-separated list. This hs no effect of sweeping is disabled.