Skip to content
This repository has been archived by the owner on Feb 5, 2022. It is now read-only.
/ Mjolnir Public archive

A tool for deploying and benchmarking Ethereum Clients

License

Notifications You must be signed in to change notification settings

brave-experiments/Mjolnir

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Mjölnir

Build Status

              |"  \    /"  |     |"  |  /    " \  |"  |    (\"   \|"  \  |" \    /"      \  
              \   \  //   |     ||  | // ____  \ ||  |    |.\\   \    | ||  |  |:        | 
              /\\  \/.    |     |:  |/  /    ) :)|:  |    |: \.   \\  | |:  |  |_____/   ) 
              |: \.        |  ___|  /(: (____/ //  \  |___ |.  \    \. | |.  |   //      /  
              |.  \    /:  | /  :|_/ )\        /  ( \_|:  \|    \    \ | /\  |\ |:  __   \  
              |___|\__/|___|(_______/  \"_____/    \_______)\___|\____\)(__\_|_)|__|  \___)

              T                                    \`.    T
                    |    T     .--------------.___________) \   |    T
                    !    |     |//////////////|___________[ ]   !  T |
                          !     `--------------'           ) (      | !
                                                     mn  '-'      !

We needed DevOps tooling to enable the Brave team rapidly deploy Ethereum Proof of Authority (PoA) clusters across different Ethereum Clients for benchmarking.

As there was no tool out there that fulfilled this requirement, this vacuum gave birth to Mjolnir ... a tool for rapidly deploying and testing Ethereum Clients.

Mjolnir allows users to test the through put of Ethereum Clients both on its own, and under adverse network conditions (i.e. Clock Skew , dropped tcp packets, jitters, etc.)

Mjolnir supports the following clients:

Table of Contents

Architecture

M

Terminology

  • {binaryName}: mjolnir
  • {client}: The Ethereum client been tested. Currently supported are:
    • quorum
    • parity
    • Support for Hyperlegder Besu is WIP.
  • {cmdName}: Binary's sub command.

Requirements

  • GNU Make (>= v3.81)
  • A UNIX based machine.
  • Go (>= v 1.12.7)
  • Docker Engine Community (>= v 19.03.1)
  • Terraform (>= v0.12.5)
  • An AWS subscription with the permissions to create IAM roles, policies, EC2 instances, s3 and Cloud watch.

In a rush?

To rapidly build , deploy and test a chain in one command, from the root folder the user should:

  • Make a copy of the yaml file i.e. cp examples/values.yml examples/values-local.yaml
  • Set the parameter for your chain and infrastructure (i.e gas limit, block time, vm type and so on)
  • From the root folder, enter make quorum-hammer
  • This should take about 20 minutes.
  • After this is done, a results folder will be created the user's root directory with the test results in it.

Getting started

  • Step 1: Deploy Infrastructure

    • Clone this repo - git clone [email protected]:brave-experiments/Mjolnir.git

    • Run

      make build

    from in root folder. This will build compile the terraform and go modules, creating a mjolnir binary in the root folder.

    • Create a copy of the configuration files in examples/values.yaml to examples/values-local.yaml i.e.

      cp examples/values.yaml examples/values-local.yaml . This is important as this file is added to the .gitignore file and protects the user from accidentally commiting secrets to their online repository.

    • Update examples/values-local.yaml with your desired configuration.

    • The user can either deploy their cluster by either of the following commands

      make {client}

      OR

      ./mjolnir apply examples/values-local.yml.

      This will deploy the requisite infrastrucure in the user's AWS account. For a 4 node cluster, this takes about 15 minutes.

  • Step 2: Fire Transactions

    • Once this is complete, enter mjolnir bastion to tunnel into the bastion host. It is from here, we are able to access chainhamer for sending transactions to the clients.
    • Move in the chainhammer directory by entering cd chainhammer
    • Run scripts/install-initialize.sh to intialize chainhammer.
    • To send transactions, CH_TXS=25000 CH_THREADING="threaded2 300" ./run.sh "{TESTNAME}"; Where CH_TXS is the number of transactions to be send, CH_THREADING is the threading algorithm, and {TESTNAME} is the name that the run will be save under.
    • If all goes well, files will be saved under:
      • ../results/runs/{client}{date}-{time}{no_of_transactions_sent}.md
      • ../results/runs/{client}{date}-{time}{no_of_transactions_sent}.html
      • ..reader/img/{TESTNAME}-{date}-{time}_blks.pgn

Build

To build from source: make build

Test

To run tests without watcher:

make test

Test-watch

To run test watcher type:

make test-watch

After a successful build , the binary mjolnir will be in the root folder.

To execute mjolnir binary file: try ./mjolnir to see all commands that are registered try ./mjolnir {cmdName} --help to see help from command

Providing values

See example/values.yml that shows how to attach values to apply execution. Since any values-local.yml file is gitignored you should copy example/values.yml to values-local.yml and provide values that you need.

In test mode cli runs with isolated scope with predefined variables and constants.

Debugging

After execution of apply command certain files will be created on your host:

  • temp.tf at root of execution dir, which contains whole terraform code that has been executed
  • terraform.tfstate at root of execution dir, which contains state of execution
  • variables.log at root of execution dir, which contains last executed variables in recipe
  • .mjolnir dir which contains necessary files like ssh key pair to bastion
  • .mjolnir/$network-name+$timestamp/ is a dir where should end private and public key pair

Build Details

To manually test build run

make build - or get latest binary release from here: https://github.com/brave-experiments/Mjolnir/releases

To execute mjolnir CLI run:

./mjolnir apply quorum values.yml or make quorum - with previosly prepared values.yml taken from examples/ folder in repo

After successful you will find following files in your working directory:

  • terraform.tfstate - current terraform object state
  • temp.tf - full dump of terraform code
  • variables.log - log file with provided vatiables

On successful run on the output you will see following example information:

Created output dir:  .mjolnir/network-name-12345678
[FINAL] Summary execution: [reset][bold][green]
Outputs:

_status = Completed!

Quorum Docker Image         = quorumengineering/quorum:2.2.5
Privacy Engine Docker Image = quorumengineering/tessera:latest
Number of Quorum Nodes      = 3
ECS Task Revision           = 1
CloudWatch Log Group        = /ecs/quorum/network-name-12345678

bastion_host_ip = xxx.xxx.xxx.xxx
bucket_name = us-east-2-ecs-network-name-12345678-b616bc76ee59e4ba
chain_id = 7774
ecs_cluster_name = quorum-network-name-12345678
ethstats_host_url = http://xxx.xxx.xxx.xxx:3000
grafana_host_url = http://xxx.xxx.xxx.xxx:3001
grafana_username = admin
grafana_password = XXXXXXXXX
network_name = network-name
private_key_file = <sensitive>
Wrote summarry output to:  .mjolnir/quorum-bastion-jkopacze-n3-66790866/output.log
Restoring env variables.

Subcommands

  • SSH into the bastion:

    ./mjolnir bastion or make bastion

  • SSH into an Ethereum node:

    ./mjolnir node n or make node n

    where n is the node number.

  • To attach an interactive geth console to any node:

    ./mjolnir geth n or make geth n

    where n is the node number.

  • Get information on currently deployed nodes

    ./mjolnir node-info or make node-info

Cleaning Up

to destroy run:

./mjolnir destroy {values-local.yml}

or

make destroy

Current success output looks like this ( will be correted in next release ):

   [FINAL] Summary execution: 
   Wrote summarry output to:  .mjolnir//output.log
   Deploy Name not present
   Restoring env variables.

Monitoring and logs

Grafana logs

For infrastructure monitoring and incident response, you no longer need to switch to other tools to debug what went wrong. Explore allows you to dig deeper into your metrics and logs to find the cause. Grafana’s new logging data source, Loki is tightly integrated into Explore and allows you to correlate metrics and logs by viewing them side-by-side More info: https://grafana.com/docs/features/explore/

To access tools to visualise your cluster's performance visit:

  • eth-stats: http://bastion_host_ip:3000
  • Grafana: http://bastion_host_ip:3001
  • Prometheus: http://bastion_host_ip:9090

Dashboard JSON

A dashboard in Grafana is represented by a JSON object, which stores metadata of its dashboard. Dashboard metadata includes dashboard properties, metadata from panels, template variables, panel queries, etc. More info: https://grafana.com/docs/reference/dashboard/

Error handling

When you are running command through CLI it should end with exit code status. Statuses are present in: commands.go

Limitations

  • This tool is meant for benchmarking alone and should not be used to deploy production instances.
  • Some features may not be compatible in Windows environment.
  • Only Amazon Web Services (AWS) is supported now. We are however open to PRs for other cloud providers!

Built with

Contributing

GIT FLOW

  • Create branch with issue you are working with, 4eg:

feature/01/CLI-initialize-project

  • If it is fix for current code that already on master:

fix/01/CLI-initialize-project-output

  • Create a Pull Pequest to branch master

  • Master is the developer branch

  • Releases will lay on certain locked branches, it will occur here after we will be ready with CI/CD pattern.

Build, deploy, run

  • All commands that must be run to fill CI/CD process are described in Makefile
  • All commands/scripts that are run from host to set up environment are run via bin/run

This logic should be sustained to clarify where code should be executed

IDE Goland

To forward go dependencies from container to your host write go mod vendor within container

Terraform Code:

  • Never use "`" (`) sign in terraform code.
  • We pass it to build as static asset, so this sign will be removed from whole string
  • All variables that are strictly used like "${var.something}" must be declared as variable in TF recipe. Otherwise validator will fail

For example:

variable "something" {
    description = "variable description"
    default = "default value"
}

Versioning

We use SemVer for versioning.

License

This project is licensed under the Mozilla Public License 2.0- see the LICENSE file for details

Frequently Asked Questions

  • Why did my deployment fail?: Although we have made efforts to provide descriptive error codes, there are times when a deployment can fail due to random faults (i.e. connectivity issues). Should this happen to you, clean up the resources and start the process again.

  • When I try to log in to the bastion, it get rejected stating that there have been too many attempts: Mjolnir registers an identity everytime you deploy new infrastructure. Unfortunately, this is not deleted after the instance is destroyed. As more as more instances are brought up, this leads to too many registers identities. Eventually when you try to log in, it fails as it iterates through all the old ones and times out. The solution is:

    • run ssh-add -l to list all the identities
    • Flush them with ssh-add -D
    • After doing this, you should be able to log into the bastion.

Acknowledgements

We would like to thank the following teams for their contributions to the project:

  • binarapps for their ability to dive into both the infrastructure and software and deliver on our requirements.
  • Dr Andreas Krueger for chainhammer. Mjolnir inspired by his project, and much of the code for firing the transactions is mostly his.
  • The JP Morgan Team for quorum-cloud. This was the boiler plate for the deployments of other clients.