cloudmesh-pi-cluster 4.3.14

A command called pi and foo for the cloudmesh shell

Documentation

• Documentation
  • About
  • Installation
    • Creating the SD Cards
    • Prerequisite
    • User install
    • Developer install
  • Using the Cluster commands
    • Getting Help
    • Setting LEDS
    • Getting the Temperature
  • Memory
    • Getting the Load Average
    • Stress test
  • Other Monitoring programs
  • Manual Pages
    • Pi Command
    • Bridge Command
  • History
    • Versions
    • Branches

This README is managed in

• README.md
• https://github.com/cloudmesh/cloudmesh-pi-cluster

About

This document describes how to set up a variety of cluster platforms on a number of Raspberry Pi's. We used Pi3B+ and Pi4 with 32 GB SD cards. You will ideally want to have a minimum of 3 Pi's.

In addition to this document, we have additional we have separate documents that showcase how to convert your cluster into a

• Spark Cluster (under development)
• Hadoop Cluster (under development)
• Kubernetes Cluster (under development)
• MongoDB Cluster (under development)

Installation

Creating the SD Cards

We have chosen not to use network booting, but boot from the SD Cards. For this, we use our unique burn program to burn the Pi's. This allows you to immediately start with an OS that has all the needed information on it. However, we need one manager Pi that we configure with the Pi imager.

TODO: point to the documentation

TODO: Briefly describe how we burn the manager and set it up

TODO: Then describe briefly how we burn

Prerequisite

Once you have set up the manager and have network access, you must conduct the following steps

First, we update the system and install python3 in ~/ENV3 with the command, activate it and generate an ssh key

curl -Ls http://cloudmesh.github.io/get/pi | sh
source ~/ENV3/bin/activate
ssh-keygen

These steps have only to be done once on your manager Pi.

User install

The installation for cloudmesh on the Pi cluster package is simple:

pip install cloudmesh-pi-cluster

Developer install

In case you like to contribute to the code, we provide a convenient mechanism for you to download all source code repositories from Git. You find in the directory where you downloaded the code all source code. The reason we recommend that you do this in a directory cm is that it is the same for all developers and that all the source code is located in one directory.

mkdir ~/cm
cd cm
cloudmesh-installer get pi

Using the Cluster commands

Getting Help

To get a list of commands related to the cloudmesh cluster use

cms help

To list the cluster command manual page use

cms help pi

to show the usage use

cms pi

Setting LEDS

In case you want to signal which Pi is which we can leverage the on board LEDs. You have a green and red LED that you can control as follows:

$ cms pi led green on
$ cms pi led green off
$ cms pi led red on
$ cms pi led red off

Getting the Temperature

Sometimes it is useful to view the temperature of the PI's as it could have performance issues due to overheating, and the PIs are forced to throttle. Hence it is a good idea to vent your Pis very well

Please note that the times are not exact as there is a slight delay between getting the data and getting the data from the Pi.

cms pi temp "red,red[01-03]"
+-------+--------+------+----------------------------+
| host  | cpu    | gpu  | date                       |
+-------+--------+------+----------------------------+
| red   | 56.965 | 56.0 | 2020-03-28 14:34:46.926618 |
| red01 | 62.322 | 62.0 | 2020-03-28 14:34:46.949065 |
| red02 | 65.731 | 65.0 | 2020-03-28 14:34:46.933548 |
| red03 | 55.017 | 54.0 | 2020-03-28 14:34:47.218047 |
+-------+--------+------+----------------------------+

Watching the temperature continuously

To watch the temperature continuously in the terminal, use a repeat rate.
To end the program press CTRL-C

cms pi temp "red,red[01-03]" --rate=1.0

Sparkline

You can also get the temperature as sparkline with

cms pi temp "red,red[01-03]"  --output=sparkline
▄▆█▃

Bar

A Bar graph is created with

cms pi temp "red,red[01-03]"  --output=bar

Line

A line graph is created with

cms pi temp "red,red[01-03]"  --output=line

Browser

The output can also be looked at in a browser

cms pi temp "red,red[01-03]"  --output=browser

Live animation

A live animation is available with

cms pi temp "red,red[01-03]" --rate=1 --output=live

Memory

cms pi free "red,red[01-03]" --rate=1

+-------+-----------+----------+----------+------------+-----------+-----------+------------+-----------+-----------+
| host  | mem.total | mem.used | mem.free | mem.shared | mem.cache | mem.avail | swap.total | swap.used | swap.free |
+-------+-----------+----------+----------+------------+-----------+-----------+------------+-----------+-----------+
| red   | 4.1 GB    | 109.5 MB | 3.8 GB   | 42.4 MB    | 188.2 MB  | 3.8 GB    | 104.9 MB   | 0 Bytes   | 104.9 MB  |
| red01 | 4.1 GB    | 99.8 MB  | 3.8 GB   | 34.0 MB    | 177.1 MB  | 3.8 GB    | 104.9 MB   | 0 Bytes   | 104.9 MB  |
| red02 | 4.1 GB    | 108.8 MB | 3.8 GB   | 34.0 MB    | 176.1 MB  | 3.8 GB    | 104.9 MB   | 0 Bytes   | 104.9 MB  |
| red03 | 4.1 GB    | 100.5 MB | 3.8 GB   | 34.0 MB    | 176.4 MB  | 3.8 GB    | 104.9 MB   | 0 Bytes   | 104.9 MB  |
+-------+-----------+----------+----------+------------+-----------+-----------+------------+-----------+-----------+

This can also be invoked repeatedly with

cms pi free "red,red[01-03]" --rate=1.0

Getting the Load Average

The load average can be obtained with

cms pi load "red,red[01-03]" --rate=1

+-------+-------+-------+------+--------------+------------+
| host  | 1     | 5     | 10   | proc.running | proc.total |
+-------+-------+-------+------+--------------+------------+
| red   | 10.01 | 10.04 | 9.58 | 1            | 142        |
| red01 | 0.01  | 0.02  | 0.0  | 1            | 125        |
| red02 | 0.03  | 0.04  | 0.01 | 1            | 128        |
| red03 | 0.09  | 0.08  | 0.02 | 1            | 125        |
+-------+-------+-------+------+--------------+------------+

We have the same formats available just as in the temperature monitor. Most useful is the live data which you can obtain with

cms pi load "red,red[01-03]" --rate=1 --output=graph

Stress test

To put some load on the system, you can use the command stress-ng

• https://manpages.ubuntu.com/manpages/artful/man1/stress-ng.1.html#examples

It can be installed on a PI with

sudo apt-get install -y stress-ng

A memory test can be started with

stress-ng --vm 8 --vm-bytes 80% -t 1h

When watching with

cms pi load "red,red[01-03]" --rate=1 --output=graph

You will see the graph reacting to it.

Other Monitoring programs

Other Monitoring programs include

mpstat -P ALL
nmon

Please make sure you install them first before using them

Manual Pages

Pi Command

  pi led reset [NAMES]
  pi led (red|green) VALUE
  pi led (red|green) VALUE NAMES [--user=USER]
  pi led list NAMES [--user=USER]
  pi led blink (red|green) NAMES [--user=USER] [--rate=SECONDS]
  pi led sequence (red|green) NAMES [--user=USER] [--rate=SECONDS]
  pi temp NAMES [--rate=RATE] [--user=USER] [--output=FORMAT]
  pi free NAMES [--rate=RATE] [--user=USER] [--output=FORMAT]
  pi load NAMES [--rate=RATE] [--user=USER] [--output=FORMAT]
  pi script list SERVICE [--details]
  pi script list SERVICE NAMES
  pi script list
  pi wifi SSID [PASSWORD] [--dryrun]

This command does some useful things.

Arguments:
    FILE   a file name

Options:
    -f      specify the file


Description:

  This command switches on and off the LEDs of the specified
  PIs. If the hostname is omitted. It is assumed that the
  code is executed on a PI and its LED are set. To list the
  PIs LED status you can use the list command

  Examples:

      cms pi led list  "red,red[01-03]"

          lists the LED status of the given hosts

      cms pi led red off  "red,red[01-03]"

          switches off the led of the given PIs

      cms pi led red on  "red,red[01-03]"

          switches on the led of the given PIs

      cms pi led red blink  "red,red[01-03]"

          switches on and off the led of the given PIs

      cms pi led red sequence  "red,red[01-03]"

          goes in sequential order and switches on and off
          the led of the given PIs

Bridge Command

  bridge create [--interface=INTERFACE] [--ip=IP] [--dns=NAMESERVER]

Options:
  --interface=INTERFACE  The interface name [default: eth1]
                         You can also specify wlan0 if you want
                         to bridge through WIFI on the manager
                         eth0 requires a USB to WIFI adapter

  --ip=IP  The ip address to assign on the eth0 interface,
           ie. the listening interface [default: 10.1.1.1]

  --dns=NAMESERVER  The ip address of a nameserver to set
                    statically. For example, --dns=8.8.8.8,8.8.4.4
                    will use the google nameservers

Description:

  Command used to set up a bride so that all nodes route the traffic
  trough the manager PI.

  bridge create [--interface=INTERFACE] [--ip=IP] [--dns=NAMESERVER]
      creates the bridge on the current device.
      A reboot is required.

History

Versions

• prebranch -- this is the code prior to us creating branches for kubernetes, and map reduce. map reduce contains spak and hadoop. However as these efforts wer just starting, they may confuse others. Thus it was important to separate them so we have a clear distinction betwen production and development.

• 02-03-2021 -- This is a branch done after prebranch, where all experimental code has been removed and placed into its seperate branches. The removal includes kubernetes, mongo, spark, hadoop. proxy and sdcard contained some documentation, that was moved to README-proxy.md and benchmar/sdcard-benchmark.md

Branches

• main: our main branch for production releases

• dev: a general branch that is used prior to the production release with active fixes

• kubernetes: a branch in which we coordinate the implemnetation of easy kubernetes deplyments

• mapreduce: a branch in which we coordinate hadoop and spark deployments. There is a temprrary solution in her, but we have not tested is and I woudl consider the status as pre alphs. It needs t o be redone.

• with-dhcp-server: Some intermediate development

• mongo: distributed deployment of mongo. This branch is essentially empty. We may delete it due to inactivity.