Skip to content

Latest commit

 

History

History
702 lines (551 loc) · 24.9 KB

README.md

File metadata and controls

702 lines (551 loc) · 24.9 KB
Raw

Table of contents

distro-pelion-edge - deb/rpm build scripts for Pelion Edge

These are scripts to build Pelion Edge as deb/rpm packages for the following distros:

  • Debian Buster (10).
  • Ubuntu Bionic (18.04).
  • Ubuntu Focal (20.04).
  • Redhat Enterprise Linux 8.
  • CentOS 8.

The folder build-env contains helper or common scripts. Other directories contain build scripts specific for each package, for Debian-based distributions:

<package name>/deb/build.sh
metapackages/<package name>/deb/build.sh

and for Red Hat-based distributions:

<package name>/rpm/build.sh

The build scripts build.sh or build-env/bin/build-all.sh can build packages directly on supported systems or in docker container. The docker container can be launched manually, for example as an interactive session, or the scripts will run docker automatically if the argument --docker is provided.

It is recommended to use docker since it will provide a clean environment without interference with local user settings (for example, with user python environments) and does not require setting up local packages repository.

Also, the build will need sudo privileges to install standard Ubuntu packages.

Quickstart

First make sure your system is configured correctly (see requirements)

Here's how to quickly build Pelion Edge Packages for Ubuntu Focal amd64.

  1. Prepare mbed_cloud_dev_credentials.c and update_default_resources.c files:
# copy your mbed_cloud_dev_credentials.c file in place so that the edge devmode package will build with your dev credentials
cp ~/Downloads/mbed_cloud_dev_credentials.c .
# use manifest tool to create an update_default_resources.c so that you'll be ready to perform an over-the-air-update
# copy the generated update_default_resources.c into place
cp /path/to/update_default_resources.c .
  1. Run build scripts:
# build all the Ubuntu 20 packages for amd64 using docker container
./build-env/bin/docker-run-env.sh focal ./build-env/bin/build-all.sh --deps --install --build --source --arch=amd64
# deb packages will be available in ./build/deploy/deb/focal/main/


# then if you'd like to rebuild an individual package (ex: mbed-edge-core-devmode)
./build-env/bin/docker-run-env.sh focal ./mbed-edge-core-devmode/deb/build.sh --install --build --source --arch=amd64

Alternative way to run build scripts:

# build all the Ubuntu 20 packages for amd64 using docker container
./build-env/bin/build-all.sh --docker=focal --deps --install --build --source --arch=amd64
# deb packages will be available in ./build/deploy/deb/focal/main/


# then if you'd like to rebuild an individual package (ex: mbed-edge-core-devmode)
./mbed-edge-core-devmode/deb/build.sh --docker=focal --install --build --source --arch=amd64

Selecting target distribution with --docker switch

The --docker switch (or short -d) is used to select build environment. Environment configurations are stored in ./build-env/target/. To list available environments run:

./build-env/bin/build-all.sh -l env

The -d accepts also partial environments names (-d rh, -d rhel/8, -d rhel-8, -d rhel - all are valid). When name matches more than one environment, script will print error:

$ ./build-env/bin/build-all.sh -d 8
Unable to load environment: ambiguous environment name, matches:
rhel/8

Preparing host to compile packages

System requirements

The build-all.sh script requires docker, bash 4.2+, gnu-findutils and gnu-getopt.

To build arm64 packages for Red Hat or Centos qemu is required and qemu-docker integration because Red Hat/Centos does not support cross compiling.

To enable qemu-docker integration on Linux binfmt has to be enabled using this image before using arm64 compilation:

docker run --rm --privileged docker/binfmt:a7996909642ee92942dcd6cff44b9b95f08dad64

The binfmt has to be executed once after every reboot.

Build on MacOS is not tested.

Red Hat Enterprise Linux (RHEL) subscription

RHEL images does not include subscription credentials. RH_USERNAME and RH_PASSWORD variables with your Red Hat subscription credentials should be exported as environment variables before running build scripts:

export RH_USERNAME="Your Red Hat username"
export RH_PASSWORD="Your Red Hat password"

Variables will be used to create docker image. If any of the variables would not be set, scripts will interactively ask for them. Credentials will be stored in docker image. Note that for free development subscription only 16 systems can be registered (registered systems can be removed on RHEL subscription management page).

To create Red Hat account visit: https://www.redhat.com/wapps/ugc/register.html

Build environment

To access build environment console docker-run-env.sh script was introduced. The script accepts options until first positional argument (which is environment name).

Example usage:

# run new rhel/8 image for host arch:
./build-env/bin/docker-run-env.sh rhel

# run centos container (exec in existing)
./build-env/bin/docker-run-env.sh -c centos

# run 'ls' in centos container
./build-env/bin/docker-run-env.sh -c centos ls

# run fresh container (and allow later executing in it)
./build-env/bin/docker-run-env.sh -c clean centos ls

# recreate docker image and container and run shell in new container
./build-env/bin/docker-run-env.sh -r -c clean centos

# run new arm64 container
./build-env/bin/docker-run-env.sh -c=clean -a arm64 centos

The above script also creates docker images with build essentials and with additional packages required for generating source packages (for example git, npm, python) depending on selected environment.

To add prefix to docker images and containers export PELION_DOCKER_PREFIX variable in your shell with your prefix:

export PELION_DOCKER_PREFIX=${USER}-

The system in docker image is configured to use sudo without a password.

The root of this repo is mounted to /pelion-build.

To get list of all supported target distributions, run:

./build-env/bin/build-all.sh -l env

It is not required to specify full name of environment (eg. ubuntu/focal). Partial, unique match would also work (like in quickstart example: focal).

Container -c and image -r flags

The -c flag enables reusing of docker containers - only one container will be used. This speeds up whole build when --docker (or -d) switch is used especially for arm64 on amd64 build. If container gets corrupted for some reason using -c=clean will create fresh container before build.

When build-all.sh is run:

  • without -c - new container is created on each element (each package build and source stage). Container is removed after each element is done:

     # create and run temporary container
 [docker run --rm] pe-utils source
 # container is automatically removed
 # create and run temporary container
 [docker run --rm] edge-proxy source
 # container is automatically removed
 ...

  • with -c script checks if there is a container, creates it if missing, and uses it for each element:

     [check if container exists => create if missing]
 [if stopped => docker start]
 [docker exec] pe-utils source
 [docker exec] edge-proxy source
 ...
 [docker exec] pe-utils build

  • with -c=clean removes existing container and creates new one before first use in script:

     # clean part
 [docker container remove]
 # create new container
 [docker container create]
 [docker start]
 [docker exec] pe-utils source
 [docker exec] edge-proxy source
 ...
 [docker exec] pe-utils build
 ...

When -c flag is used in build-all.sh, the container can be accessed with docker-run-env.sh -c command. It can be run multiple times so multiple shell sessions can be created in one container. It also allow to reboot host machine and attach to previously used session.

As script now automatically creates required images, -r flag was introduced. The -r flag forces script to recreate docker images.

Building a single package

Dependency packages must be built prior to building a single package.

./build-env/bin/build-all.sh --deps --install --docker=<dist>

or

./build-env/bin/docker-run-env.sh <dist> ./build-env/bin/build-all.sh --deps --install

Build single package for Debian and Ubuntu

The build scripts provide help information, for example:

$ pe-utils/deb/build.sh --help
Usage: pe-utils/deb/build.sh [Options]

Options:
 --docker            Use docker containers.
 --source            Generate source package.
 --build             Build binary from source generated with --source option.
 --verify            Verify package conformity to the Debian policy.
 --install           Install build dependencies.
 --arch=<arch>       Set target architecture.
 --print-target      Print target package file path and exit
 --print-package-name Print package name (eg. devicejs) and exit
 --print-package-version Print package version (eg. devicejs) and exit
 --help,-h           Print this message.

 If none of '--source', '--build' or '--verify' options are specified,
 all of them are activated.

Available architectures:
  amd64
  arm64
  armhf
  armel

Default mode: pe-utils/deb/build.sh --arch=amd64 --source --build --verify

These scripts can be used to generate both source and binary packages.

It is possible to use these scripts without the option --docker if docker is run manually by docker-run-env.sh. The option --docker will make all tasks run in appropriate containers automatically (using source of build images):

Build single package for Red Hat and Centos

The build scripts for RPM-based distributions are similar to Debian's:

$ pe-utils/rpm/build.sh --help
build.sh - builds pe-utils RPM package

Usage: build.sh [-h|--help] [--install]
 -h, --help                 display this help message.
 -i, --install              install dependencies
 -b, --build                build package
 -s, --source               prepare sources
 -d, --docker=<name>        use docker container to build RPM
 -a, --arch=<name>          build for selected architecture
 -c, --container=[opts]     reuse container; opts can be (comma separated):
                            - clean - create new container before first use
 -r, --recreate             forcibly recreate docker images (implies -c=clean)
 -o, --deploy=<path>        set target directory for RPMs

Additionally these scripts supports Docker image creation when needed and container re-usage.

Building all packages

$ ./build-env/bin/build-all.sh --help
build-all.sh - build all packages:

USAGE: 
-a, --arch=<name>       run build for <name> architecture
-d, --docker=<name>     run build in docker (name=docker environment)
-h, --help              print this help text
-i, --install           install dependencies
-e, --print-env=[what]  print environment setup: packages, meta, deps
                        (deps packages)
-l, --print-list        print list: env (list of available environments)
-c, --container=[opts]  use one container per kind instead of container
                        per package; opts can be (comma separated):
                        - clean - create new container before first use
-r, --recreate          forcibly recreate docker images (implies -c=clean)

Build elements:
-b, --build             run build stage
-s, --source            run source stage
-t, --tar               run tarball creation stage
-p, --deps              run dependency compilation stage

When no -b, -s, -t or -p are set, all are enabled. Setting one of them will
disable unset parameters (eg. setting -b -s will run only source and build
stage).

If --docker is set, build process will run in new container. By adding --container
scripts will use one container to build all packages (instead of container
per package)

This script will run build of each package in new docker container installing all build dependencies each time (with --docker option), for example:

./build-env/bin/build-all.sh --docker=focal --arch=amd64,armhf,arm64

By adding --container option, one container will be used if possible. Adding --container clean will remove existing container before build.

It is possible to manually run docker and then build everything in this container:

$ ./build-env/bin/docker-run-env.sh bionic
user@95a30883d637:/pelion-build$ ./build-env/bin/build-all.sh --install --arch=amd64

When build requires different architecture (when cross-compilation is not available, like for Red Hat) --arch=<architecture> has to be added to docker-run-env.sh before specifying target environment:

./build-env/bin/docker-run-env.sh --arch arm64 rhel

The option --install is required when the script is executed in new docker container, otherwise it will fail due to missing build dependencies.

To provide build dependency use --deps switch. This will create build dependencies and put into local repository. This is enabled by default if not build/source/tar is set.

Generating tar archives

A tar archive can be created from a binary release for Debian and Ubuntu only. One option is to use the build-all.sh script as described above. Another option is to invoke build-env/bin/deb2tar.sh directly.

$ build-env/bin/deb2tar.sh --help
deb2tar.sh - converts a set of Debian packages into a portable tarball.

Usage: deb2tar.sh [-h|--help] [-a ARCH|--arch ARCH]
 -h, --help
  Display this help message.
 -a ARCH, --arch ARCH		Set the host architecture of the tarball.
 -d DISTRO, --distro DISTRO Set Linux distro (eg. focal, buster...)

Before invoking deb2tar.sh make sure that all packages you want to be included in the tarball are already built.

Build results

The scripts create or modify files only in the build directory that is also created automatically.

There is a reusable git cache in the directory build/downloads.

Final Debian packages are created in build/deploy/deb/<DISTRO>/main, organized in subdirectories per architecture:

  • binary-amd64
  • binary-arm64
  • binary-armhf
  • source

Tarballs can be found in build/deploy/tar, one archive per architecture.

Red Hat and Centos packages are created in build/deploy/rpm/<DISTRO>:

  • source packages are on top of this directory
  • noarch, x86_64 and aarch64 directories contains final packages

Preparing Red Hat system before installation

Here are notes for installation compiled packages on target system.

RHEL Repositories

CodeReady and EPEL have to be enabled before installation EPEL is required only for mbed-edge-example package.

sudo subscription-manager repos --enable codeready-builder-for-rhel-8-x86_64-rpms
sudo yum install -y https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm

Docker

RHEL 8 does not support Docker, to use kubelet it is required to install Docker from external repository. This is required only to run built binaries (not required to do the actual build). To add the repository, run:

sudo yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo

SELinux

SELinux is not currently supported: pelion-relay-term service will be killed when SELinux is enabled.

SELinux has to be disabled, set to permissive or node binary should be excluded. More details about SELinux and its configuration here: https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/using_selinux/index

Installing packages

Build results can be installed onto a target system either by manually copying the packages to a target system and installing via apt/yum or by making the packages available in an package repository on a server and installing on the target via apt/yum. The instructions in this section show how to install the packages manually. See the next section for setting up an APT repository (Debian/Ubuntu only).

Installing on Debian or Ubuntu

Copy the debian packages found in build/deploy/deb/<DISTRO> to the target system and install with apt.

Install all packages with the following command.

$ sudo apt install -y ./*.deb

Or, install a single package by specifying its deb file name.

$ sudo apt install -y ./edge-proxy<version>_<arch>.deb

Additional steps for Debian 11 Bullseye

Pelion container orchestration services require cgroupv1 for proper operation. Since Debian 11 changes to using cgroupv2 by default, the Debian 11 kernel must be configured to override the default and use a hybrid cgroup heirarchy. Do this by adding the parameters systemd.unified_cgroup_hierarchy=false and systemd.legacy_systemd_cgroup_controller=false to the kernel command line.

For example, if using grub, add the following line to a new file /etc/default/grub.d/pelion.cfg: GRUB_CMDLINE_LINUX_DEFAULT="${GRUB_CMDLINE_LINUX_DEFAULT} systemd.unified_cgroup_hierarchy=false systemd.legacy_systemd_cgroup_controller=false" Next, regenerate the grub config file and reboot the system. sudo update-grub sudo reboot

Installing on Red Hat or Centos

  1. Before installing packages make sure that you have subscription enabled and EPEL, CodeReady and Docker repositories are enabled (see RHEL repositories and Docker).

  2. Copy content of build/deploy/rpm/<distro>/<arch> and build/deploy/rpm/<distro>/noarch/ to target system (where <arch> is amd64 or arm64 and <distro> is rhel8 ).

  3. To install use yum command, for example if all packages are in current directory run:

sudo yum install *.rpm

Please note that mbed-edge-core, mbed-edge-core-devmode, and mbed-edge-core-byocmode cannot be installed simultaneously.

  1. Enable systemd services. After installation there are following services:
edge-core.service
edge-proxy.service
kubelet.service
edge-resource-manager.service
mbed-fcc.service
pe-terminal.service
wait-for-pelion-identity.service
fluent-bit.service

To enable all services, run:

sudo systemctl enable edge-core.service edge-proxy.service kubelet.service edge-resource-manager.service mbed-fcc.service pe-terminal.service wait-for-pelion-identity.service

Dependent services are enabled implicitly. For example wait-for-pelion-identity.service is enabled when edge-resource-manager.service is enabled; edge-core.service is enabled when wait-for-pelion-identity.service is enabled so enabling edge-resource-manager.service will also enable wait-for-pelion-identity.service and edge-core.service.

Removing packages

List of packages can be printed with following command (here example for packages in Ubuntu Focal):

$ ./build-env/bin/build-all.sh -d focal -l packages

To remove package on Debian/Ubuntu use command:

$ sudo apt remove -y <package name> --autoremove --purge

or on Red Hat/Centos:

$ sudo yum remove -y <package name>

After removing all packages, manually remove credentials and config files:

$ sudo rm -rf /var/lib/pelion/
$ sudo rm -rf /etc/pelion/

APT repository

Structure for APT repository server can be created in build/deploy/deb/apt-repo directory by build-env/bin/pelion-apt-repo-create.sh. It provides help information:

$ ./build-env/bin/pelion-apt-repo-create.sh --help
Usage: pelion-apt-repo-create.sh [Options]

Options:
 --key-name=<name>         Filename of secret GPG key.
 --key-[id|path]=<id|path> Use key id of existing GPG key or path where private key is placed.
 --install                 Installs the necessary tools to create structure for apt repository.
 --help,-h                 Print this message.

An APT repository and its contents have to be signed with a GPG key. There are two ways to pass such a key to the script: using an id of a key stored in one's GPG keyring or a filesystem path to a secret key. The default is to use the file build/deploy/deb/gpg/Pelion_GPG_key_private.gpg.

GPG key pair generation

To generate a key pair one can use build-env/bin/pelion-gpg-key-generate.sh:

$ ./build-env/bin/pelion-gpg-key-generate.sh --help
Usage: pelion-gpg-key-generate.sh --key-email=<email> [Options]

Options:
 --key-name=<name>             Set name of GPG key and filename of public (<name>_public.gpg)
                               and private (<name>_private.gpg) keys.
 --key-email=<email>           Set email of GPG key.
 --key-path=<path>             Set path where public and private keys will be placed.
 --install                     Installs the necessary tools to generate the gpg key pair.
 --help,-h                     Print this message.

The --key-email flag is required (it's the email part of a key's id). The generated public and private keys will be placed into build/deb/deploy/gpg directory by default, but this can be changed using the --key-path option.

APT repository usage

In order to get access to an apt repository add a line of the following form in /etc/apt/sources.list:

deb [arch=amd64] http://<ip address> bionic main

Then import the GPG key the repository was signed with:

wget -q -O - http://<ip address>/key.gpg | sudo apt-key add -

YUM repository

Structure for YUM repository server can be created in build/deploy/rpm/yum-repo. createrepo tool needs to be installed to create repository. Package files need to be copied to this location and then repository can be created with createrepo tool.

$ mkdir -p build/deploy/rpm/yum-repo
$ cp <packages path> build/deploy/rpm/yum-repo

$ yum install createrepo
$ createrepo build/deploy/rpm/yum-repo

$ yum clean all
$ yum update

An YUM repository and its contents have to be signed with a GPG key. Key can be generated using this instruction and will be placed here build/deploy/deb/gpg/Pelion_GPG_key_private.gpg. Export this key to file with gpg tool and import key to rpm tool to be able to sign packages.

$ gpg --export -a 'Pelion_GPG_key' > RPM-GPG_KEY-pelion
$ rpm --import RPM-GPG_KEY-pelion

In order to utilize the key edit the file ~/.rpmmacros.

%_signature gpg
%_gpg_path <user's home directory path>/.gnupg
%_gpg_name Pelion_GPG_key
%_gpgbin /usr/bin/gpg

Now all packages can be signed with GPG key.

$ rpm --addsign build/deploy/rpm/yum-repo/*.rpm

YUM repository usage

In order to get access to an YUM repository add a new file in this path /etc/yum.repos.d/pelionedge.repo with this contents:

[pelionedge]
name=pelionedge
baseurl=http://<ip address>
enabled=1
gpgcheck=1
gpgkey=file://<RPM-GPG_KEY-pelion (public) key path>

gpgkey field can also be remote web link.

FOTA

Firmware Over-The-Air(FOTA) behavior is currently undefined

Known Issues

iptables versioning

There is a bug in iptables versions 1.8.0->1.8.2 that causes kubelet to create duplicate firewall rules.

For more information, see the following known issues:

Some suggested workarounds include:

  • Running the following command on the host: update-alternatives --set iptables /usr/sbin/iptables-legacy
  • Upgrade iptable to version 1.8.3+