Kubernetes deployments and stateful sets scaler via slack bot slash command

diagram

Kubernetes cluster with multiple namespaces. These namespaces should be stopped at specific time every working day and started manually by user request.

requirements

Namespace should be stopped by cronjob at specific time if it has scheduler=enabled label. All namespaces should have the same scheduled shutdown time (for now it's hardcoded in template and can be found in src/templates/cronjob.py).

cronjob is created automatically and should be active when:

• we create a new namespace without label
• we label a namespace with scheduler=enabled

cronjob should be paused when:

• namespace has been created with scheduler=disabled
• namespace has been relabeled with scheduler=disabled

Additional cronjob should reset shutdown time in all namespaces to default one every morning.

CREATE and UPDATE kubernetes events on namespace should be managed by AdmissionWebHook. (see app-scaler/controllers/cronjob_controller.py`).

additional requirements

EC2 or RDS instances should be stopped or started if they have the same labels. For instance,

namespace: dev1
scheduler: enabled

We should have a possibility to run static pod for management purpose. Request should consist of memory size of a pod (like XS,S,M etc).

There should be at least 3 types of notifications before environment is going to shutdown: 1 hour before, 30 and 15 mins before shutdown.

slack slash commands

Commands should send default payload to scaler api; that means it will scale all deployments and stateful sets to 0 replicas as stop/start commands.

prolongate

Namespace can be prolongated: user can modify (add hours) shutdown time by invoking command / prolongate for . Example:

/devenv prolongate dev1 for 1h

start

Namespace can be started manually by user request by invoking command / start . Example:

/devenv start dev1

stop

Namespace can be stopped manually by user request by invoking command / stop . Example:

/devenv stop dev1

status

We can track actual status of specific deployment (based on kubernetes deployment status) via / status . Example:

/devenv status dev1

help

Slash command should have help command which will return actual help message for all commands via / help. Example:

/devenv help

suppoer

Slack command to start and stop manage pod. / support start

/devenv support start manage1 XL
/devenv support stop manage1

scaler api

OpenAPI scheme is on /docs url.

cacher

Purpose is to save ec2 and rds instances with provided labels into dynamodb as cache location. Written in Golang.

proxy

A proxy for scaler service because AdmissionWebhook needs a ssl endpoint.