Skip to content

Latest commit

 

History

History
92 lines (78 loc) · 4.98 KB

INFO.md

File metadata and controls

92 lines (78 loc) · 4.98 KB
Raw

Prerequisites:

General info

This Terraform is made for weka deployment on GCP including auto-scaling. This Terraform can use existing network (vpcs/subnets etc.) or create new network.

On our root module you will find the actual weka deployment. It includes instance template, cloud functions, workflows, job schedulers, secret manger, buckets, health check

We supply additional 4 modules:

  1. setup_network: includes vpcs, subnets, peering, firewall and health check.
  2. service_account: includes the service account that will be used for deployment with all necessary permissions.
  3. worker_pool: includes build private pool creation.
  4. shared_vpcs(optional): includes vpc sharing between the weka deployment network and another notwork.

We support deploying weka on public and private network.

  • public network deployment:
    • requires passing get.weka.io token to terraform.
  • private network deployment:
    • requires weka installation tar file in some GCP bucket
    • utils folder contains helper script utils/sync_weka_tar.sh

You can find several usage examples under examples in this repo.

After applying this Terraform, you will get 2 workflows that run every minute and will be used for scale up and down. Automatically a new cluster will be created in a few minutes according to the cluster size you set. When an instance is added to the instance group the Terraform created, it indicates this instance is now a part of the cluster.

In order to change the cluster size (up or down) you can use a special cloud function we made called resize. Example:

curl -m 70 -X POST RESIZE_CLOUD_FUNCTION_URL -H "Authorization:bearer $(gcloud auth print-identity-token)" -H "Content-Type:application/json" -d '{"value":6}'

Similar command

Weka cluster deployment:

  • Authentication: gcloud auth application-default login
  • Deployment:
    This repository provides TF modules and examples, actual usage assumes use of modules in new or existing project, similar way as shown in examples. Examples include examples for public and private networks Important to note:
    • Public deployment requires to pass get_weka_io_token in order to download release from public get.weka.io service
    • Private deployment requires to upload weka software tarfile into GSC bucket, so instances will be able to download software from it
    • In addition, following params are optional for private networking, depending on how network topology looks like:
      • install_weka_url on deploy_weka module level, this allows to download weka from local bucket and not public get.weka.io service
      • yum_repo_server - Centos7 only, instructions to auto-configure yum to use alternative repository. Distributive repository required in order to download kernel headers and additional build software
      • source_image_id - custom image to use

Notes

  • You have 2 ways to know that your weka cluster is ready:
    • all vms where added to the instance group
    • run resize curl command with target size equal to initial size. Until cluster is fully formed, it will return error indication that cluster is not ready yet
    • future versions will include status api with more information
  • In case you deployed a public cluster you can't change it to private and vise versa
  • You can't change vpc or number of nics after deployment
  • In order to see the input and output of each step in the scale down workflow, you can go to EDIT, then you can edit the scheduler, go to Configure the execution and choose for log level All calls . (We can't set this option via TF)
  • Right now only two configurations are supported:
    • nics_number == 4 with instance type c2-standard-8
    • nics_number == 7 with instance type c2-standard-16

Support Regions

GCP not support function on the regions below, mapping added:

europe-west4 = "europe-west1" europe-north1 = "europe-west1", us-east5 = "us-east1", southamerica-west1 = "northamerica-northeast1", asia-south2 = "asia-south1",

Map function resources:

  • vpc_connector - map can be found at /modules/setup_network/variables.tf using variable vpc_connector_region_map
  • cloud_scheduler - map can be found at /modules/deploy_weka/variables.tf using variable cloud_scheduler_region_map
  • cloud_functions - map can be found at /modules/deploy_weka/variables.tf using variable cloud_functions_region_map

GCP not support workflow at region southamerica-west1, mapping added:

southamerica-west1 = "southamerica-east1"

Map workflow resource:

  • workflow - map can be found at /modules/deploy_weka/variables.tf using variable workflow_map_region

Regions are not support for weka (machine type c2 is not available)

  • europe-central2
  • northamerica-northeast2
  • us-south1
  • asia-southeast2
  • australia-southeast2
  • europe-southwest1
  • europe-west8
  • europe-west9