Giter Site home page Giter Site logo

nbxorg / mountain-gorilla Goto Github PK

View Code? Open in Web Editor NEW

This project forked from tritondatacenter/mountain-gorilla

0.0 2.0 0.0 5.22 MB

Builder of all the SDC bits.

License: Mozilla Public License 2.0

Makefile 45.85% Shell 30.40% JavaScript 22.32% Python 1.42%

mountain-gorilla's Introduction

mountain-gorilla

This repository is part of the Joyent Triton project. See the contribution guidelines -- Triton does not use GitHub PRs -- and general documentation at the main Triton project page.

A single repo to build all the parts of Triton. This is just a build driver repo, all the components are still in their respective repos. See https://mo.joyent.com/docs/mg for a more complete introduction.

Quick start

While MG theoretically knows how to "build the world", i.e all of Triton, the typical usage is to build one piece at a time. There is a make target (or targets) for each Triton component. So, for example, here is how you build VMAPI:

git clone --origin=cr https://cr.joyent.us/joyent/mountain-gorilla.git
cd mountain-gorilla
./configure -t vmapi -d Joyent_Dev  # generates bits/config.mk and fetches repo and deps
make vmapi                          # builds in build/vmapi, bits in bits/...

If that fails for you, you might be missing prerequisites. See https://mo.joyent.com/docs/mg/master/#prerequisites.

If you'll actually be building, see "Prerequisites" section below first.

The "bits/config.mk" contains all config information necessary to fully reproduce the build. There will be configure options to use some prebuilt bits (e.g. a prebuilt platform) -- so to fully reproduce in this scenario those pre-built bits will need to be available. The "configure" step might take a while because it needs to get/update all of the source repositories to get revision information (the git shas to be built are part of the created "config.mk" file)

The end result is a "bits" directory with all the built components. Primarily this includes the release bits in "bits/usbheadnode": "coal-$VERSION.tgz", "usb-$VERSION.tgz", "boot-$VERSION.tgz" and "upgrade-$VERSION.tgz". However, also included are all the constituent built bits: agents, platform, ca, etc.

The above configuration is to build the world from scratch. That takes around 2 hours. You can also build just the individual pieces, e.g. cloudapi:

./configure -t cloudapi
make cloudapi

Likewise for any target (cat targets.json | json --keys)

To include ancillary closed repositories in a usb-headnode build (which the 'usb-headnode' and 'usb-headnode-debug' targets will build), you need to pass the -j option to configure. Note that this will cause your build to fail if you do not have access to the private ancillary repositories, so it should be used only when building Joyent products.

Prerequisites

You need several components to build our images in Triton. The easiest way to get these is to use the (private repo) jenkins-agent builds but you can also manually install all the things from those images.

You should now be able to build mountain-gorilla (MG): i.e. all of Triton. Let's try that:

git clone [email protected]:joyent/mountain-gorilla.git
cd mountain-gorilla
time (./configure && gmake) >build.log 2>&1 &; tail -f build.log

Adding a repository quickstart

Add it as a top-lever property in targets.json, as an object with properties "repos" and "deps" minimally, both are arrays.

  • "repos" is an array of objects, with the property "url", pointing at a git url
  • "deps" is an array of strings, where the string is another top-level target in targets.json (or optionally a "$branch/$target" to lock it to a branch, though the need for this should be rare).

For example:

{
  ...
  mynewrepo: {
    "repos": [ {"url": "git://github.com/joyent/mynewrepo.git" } ],
    "deps": [ "platform" ]
  },
  ...
}

Then you'll add the target to Makefile. MG's configure will automatically populate some Makefile values for you, noteably: xxx_BRANCH , xxx_SHA, but you will need to fill in the build stamp yourself. Configure will also git checkout your repo in build/

#---- MYNEWREPO

_mynewrepo_stamp=$(MYNEWREPO_BRANCH)-$(TIMESTAMP)-g$(MYNEWREPO_SHA)
MYNEWREPO_BITS=$(BITS_DIR)/mynewrepo/mynewrepo-pkg-$(_mynewrepo_stamp).tar.bz2

.PHONY: mynewrepo
mynewrepo: $(MYNEWREPO_BITS)

$(mynewrepo_BITS): build/mynewrepo
  mkdir -p $(BITS_DIR)
  (cd build/mynewrepo && TIMESTAMP=$(TIMESTAMP) BITS_DIR=$(BITS_DIR) gmake pkg release publish)
  @echo "# Created mynewrepo bits (time `date -u +%Y%m%dT%H%M%SZ`):"
  @ls -1 $(MYNEWREPO_BITS)
  @echo ""

clean_mynewrepo:
  rm -rf $(BITS_DIR)/mynewrepo
  (cd build/mynewrepo && gmake clean)

if you wish to build an application zone image, the process is roughly similar except you will need to add the "appliance":"true" property, the "pkgsrc" property and "dataset_uuid"

{
  ...
  "mynewrepo": {
    "repos" : [ {"url":"git://github.com/joyent/mynewrepo.git"} ],
    "appliance": "true",
    "dataset_uuid": "01b2c898-945f-11e1-a523-af1afbe22822",
    "pkgsrc": [
      "sun-jre6-6.0.26",
      "zookeeper-client-3.4.3",
      "zookeeper-server-3.4.3"
    ],
    deps: []
  },
  ...
}

where dataset_uuid is the uuid of the source image you wish to build off pkgsrc is an array of strings of package names to install.

Your Makefile target will look as above, with the addition of the xxx_dataset target:

...
MYNEWREPO_DATASET=$(BITS_DIR)/mynewrepo/mynewrepo-zfs-$(_mynewrepo_stamp).zfs.bz2

.PHONY: mynewrepo_dataset

mynewrepo_dataset: $(MYNEWREPO_DATASET)

$(MYNEWREPO_DATASET): $(MYNEWREPO_BITS)
        @echo "# Build mynewrepo dataset: branch $(MYNEWREPO_BRANCH), sha $(MYNEWREPO_SHA), time `date -u +%Y%m%dT%H%M%SZ`"
        ./tools/prep_dataset.sh -t $(MYNEWREPO_BITS) -o $(MYNEWREPO_DATASET) -p $(MYNEWREPO_PKGSRC)
        @echo "# Created mynewrepo dataset (time `date -u +%Y%m%dT%H%M%SZ`):"
        @ls -1 $(MYNEWREPO_DATASET)
        @echo ""
...

prep_dataset.sh is a script that generates images out of tarballs and lists of packages.

It takes arguments of the form -t where is a .tar.gz file, containing a directory "root", which is unpacked to / -p "list of pkgsrc packages" where list of pkgsrc packages is a list of the pkgsrc packages to be installed in the zone.

Configure will populate xxx_DATASET and xxx_PKGSRC based on targets.json.

Additionally, you can set the dsadm URN for the target by adding the "urn" and "version" properties to targets.json, as properties of the target you wish to manipulate. These will show up as urn:version ( sdc:sdc:mynewrepo:0.1 for instance ). To use them, configure will populate xxx_URN and xxx_VERSION for you in the Makefile.

Note that these images can only be provisioned with the joyent-minimal brand. If one is provisioned with the joyent brand, that zone's networking may not be working. Normally, the networking setup is done through zoneinit, but since that script has already run and had its effects undone (as part of the MG build), there's no mechanism to automatically bring that zone's VNIC up. You can recover by manually enabling network/physical:default, but you should just be provisioning with the joyent-minimal brand instead. See RELENG-337 for details.

Package Versioning

Thou shalt name thy Triton constituent build bits as follows:

NAME-BRANCH-TIMESTAMP[-GITDESCRIBE].TGZ

where:

  • NAME is the package name, e.g. "smartlogin", "ca-pkg".

  • BRANCH is the git branch, e.g. "master", "release-20110714". Use:

      BRANCH=$(shell git symbolic-ref HEAD | awk -F / '{print $$3}')  # Makefile
      BRANCH=$(git symbolic-ref HEAD | awk -F / '{print $3}')         # Bash script
    
  • TIMESTAMP is an ISO timestamp like "20110729T063329Z". Use:

      TIMESTAMP=$(shell TZ=UTC date "+%Y%m%dT%H%M%SZ")    # Makefile
      TIMESTAMP=$(TZ=UTC date "+%Y%m%dT%H%M%SZ")          # Bash script
    

    Good. A timestamp is helpful (and in this position in the package name) because: (a) it often helps to know approx. when a package was built when debugging; and (b) it ensures that simple lexographical sorting of "NAME-BRANCH-*" packages in a directory (as done by agents-installer and usb-headnode) will make choosing "the latest" possible.

    Bad. A timestamp sucks because successive builds in a dev tree will get a new timestamp: defeating Makefile dependency attempts to avoid rebuilding. Note that the TIMESTAMP is only necessary for released/published packages, so for projects that care (e.g. ca), the TIMESTAMP can just be added for release.

  • GITDESCRIBE gives the git sha for the repo and whether the repo was dirty (had local changes) when it was built, e.g. "gfa1afe1-dirty", "gbadf00d". Use:

      # Need GNU awk for multi-char arg to "-F".
      AWK=$((which gawk 2>/dev/null | grep -v "^no ") || which awk)
      # In Bash:
      GITDESCRIBE=g$(git describe --all --long --dirty | ${AWK} -F'-g' '{print $NF}')
      # In a Makefile:
      GITDESCRIBE=g$(shell git describe --all --long --dirty | $(AWK) -F'-g' '{print $$NF}')
    

    Notes: "--all" allows this to work on a repo with no tags. "--long" ensures we always get the "sha" part even if on a tag. We strip off the head/tag part because we don't reliably use release tags in all our repos, so the results can be misleading in package names. E.g., this was the smartlogin package for the Lime release:

      smartlogin-release-20110714-20110714T170222Z-20110414-2-g07e9e4f.tgz
    

    The "20110414" there is an old old tag because tags aren't being added to smart-login.git anymore.

    "GITDESCRIBE" is optional. However, the only reason I currently see to exclude it is if the downstream user of the package cannot handle it in the package name. The "--dirty" flag is optional (though strongly suggested) to allow repos to deal with possibly intractable issues (e.g. a git submodule that has local changes as part of the build that can't be resolved, at least not resolved quickly).

  • TGZ is a catch-all for whatever the package format is. E.g.: ".tgz", ".sh" (shar), ".md5sum", ".tar.bz2".

Exceptions

The agents shar is a subtle exception:

agents-release-20110714-20110726T230725Z.sh

That "release-20110714" really refers to the branch used to build the agent packages included in the shar. For typical release builds, however, the "agents-installer.git" repo is always also on a branch of the same name so there shouldn't be a mismatch.

Suggested Versioning Usage

It is suggested that the Triton repos use something like this at the top of their Makefile to handle package naming (using the Joyent Engineering Guidelines, eng.git):

include ./Makefile.defs   # provides "STAMP"
...
PKG_NAME=$(NAME)-$(STAMP).tgz

Notes:

  • This gives the option of the TIMESTAMP being passed in. This is important to allow an external driver -- e.g., moutain-gorilla, bamboo, CI -- to predict the expected output files, and hence be able to raise errors if expected files are not generated.
  • Consistency here will help avoid confusion, and surprises in things like subtle differences in awk on Mac vs. SmartOS, various options to git describe.

mountain-gorilla's People

Contributors

trentm avatar johnsonnenschein avatar joshwilsdon avatar orlandov avatar kusor avatar jclulow avatar fkuo avatar rmustacc avatar melloc avatar mcavage avatar yunong avatar pfmooney avatar cburroughs avatar arekinath avatar nshalman avatar jjelinek avatar johananlai avatar thekvn avatar dwlf avatar twhiteman avatar 13w avatar cjihrig avatar

Watchers

Netbox avatar James Cloos avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.