Initial how to build info and Make targets #107

ashcrow · 2018-06-15T13:52:58Z

A question has come up over how one would build RHCOS. The original Container Linux has this document explaining how to do so. Today the closest we have is for one to look at Jenkinsfile.*s.

This would let people run one or more make targets locally to build a local development version of RHCOS.

See #108

jlebon · 2018-06-15T13:58:08Z

README-development.md

+- ostree
+- rpm-ostree 
+- imagefactory-plugins-TinMan
+- qemu-img


Hmm, might be worth inlining these requires into the building steps. That way if one just wants to know how to compose an OSTree, they don't need to install e.g. imgfac (which pulls in libvirt + systemd).

Offline suggestion: link the exotic tools to their repo (or, if they have it, to their "how to install" doc).

Most of this is yum -y install, with a few exceptions, those we should call out.

jlebon · 2018-06-15T13:59:28Z

README-development.md

+ - ``ostree --repo=repo remote add rhcos --no-gpg-verify ${OSTREE_INSTALL_URL}``
+ - ``ostree --repo=repo pull --mirror --commit-metadata-only rhcos``
+ - ``imagefactory --debug base_image --file-parameter install_script cloud.ks --parameter offline_icicle True rhcos.tdl``
+ - ``export IMAGE=`ls /var/lib/imagefactory/storage/*.body```


Oh neat! I didn't know that you could use double ticks on the outside to safely use single ticks ` inside them!

I'm not sure you can 😆

cgwalters · 2018-06-15T17:27:57Z

README-development.md

@@ -1,3 +1,39 @@
+# Operating System


I run basically all of this from a container, and so does the pipeline (except for Imagefactory for bad reasons). I think we should recommend that as well; it's mentioned below for the rpm-ostree bits. In fact we could just recommend pulling quay.io/cgwalters/coreos-assembler and running things in there right?

Switched coreos-assembler to use podman

cgwalters · 2018-06-15T17:28:22Z

README-development.md

+- ostree
+- rpm-ostree 
+- imagefactory-plugins-TinMan
+- qemu-img


Most of this is yum -y install, with a few exceptions, those we should call out.

cgwalters · 2018-06-15T17:29:37Z

README-development.md

+ - ``rpmdistro-gitoverlay build``
+- Make the ostree
+ - ``export REPO=/tmp/repo``
+ - ``ostree init --repo=$REPO --mode=archive --src $REPO/ --dest=$REPO``


These --src/--dest args seem like a copy/paste error from rsync?

I actually wasn't 100% sure the right way to convert these lines from the Jenkinsfile without actually going through and testing out to see what happens.

It doesn't look like --src $REPO/ --dest=$REPO was dropped, definitely should be.

It's now removed

cgwalters · 2018-06-15T17:29:43Z

README-development.md

+ - ``export REPO=/tmp/repo``
+ - ``ostree init --repo=$REPO --mode=archive --src $REPO/ --dest=$REPO``
+ - ``coreos-assembler --repo=$REPO host.yaml``
+ - ``ostree summary --repo=$REPO --update --dest $REPO/ --src=$REPO``


cgwalters · 2018-06-15T17:30:11Z

README-development.md

+- Make the ostree
+ - ``export REPO=/tmp/repo``
+ - ``ostree init --repo=$REPO --mode=archive --src $REPO/ --dest=$REPO``
+ - ``coreos-assembler --repo=$REPO host.yaml``


rpm-ostree from git master supports the yaml...but this is OK for now.

jlebon · 2018-06-15T18:51:31Z

README-development.md

+- Make the qcow2
+ - ``ostree --repo=repo remote add rhcos --no-gpg-verify ${OSTREE_INSTALL_URL}``
+ - ``ostree --repo=repo pull --mirror --commit-metadata-only rhcos``
+ - ``imagefactory --debug base_image --file-parameter install_script cloud.ks --parameter offline_icicle True rhcos.tdl``


This is missing two seds for subbing out OSTREE_INSTALL_URL and OSTREE_INSTALL_REF. Though we can probably hardcode the latter at least now. (It was originally written to be usable for both CentOS and RHEL-derived input).

lucab · 2018-06-18T13:36:40Z

README-development.md

+
+- git
+- coreos-assembler
+- rpmdistro-gitoverlay


This additionally requires fedpkg and mock.

ashcrow · 2018-06-18T13:51:14Z

Updates based on feedback.

miabbott · 2018-06-18T16:12:12Z

README-development.md

+ - ``export REPO=/tmp/repo``
+ - ``ostree init --repo=$REPO --mode=archive``
+ - ``podman run --privileged --rm -v $REPO:/srv quay.io/cgwalters/coreos-assembler``
+ - ``cd /srv && coreos-assembler --repo=/srv --cachedir=cache host.yml``


I used the coreos-assembler container to run all the commands up to here (including the rdgo ones).

If you follow the steps verbatim, the coreos-assembler command will fall over because there is no openshift.repo file available in openshift/os

I was able to generate the openshift.repo with the generate-openshift-repo command in the openshift/os repo, but then we fall over due to missing the Red Hat GPG key.

So I propose that the additional steps of:

running generate-openshift-repo

copying the Red Hat GPG key into /etc/pki/rpm-gpg/

...then running the coreos-assembler command.

(It might also be worth noting separately that the coreos-assembler command is not going to work unless you have access to the Red Hat stage CDN, just to dissuade anyone who is curious)

I was also going through the steps, and hit the same. For the former there seems to be already a make repo-refresh (similarly for init). For the latter I'm confused because it seems to bail on a gpgcheck=0 repo; is there a way to hook repo configs to relative paths from this git-repo?

Yeah, we can.

cgwalters · 2018-06-19T14:30:03Z

README-development.md

+- mock
+
+### May need to be built from source
+- [coreos-assembler](https://github.com/cgwalters/coreos-assembler/): [container](http://quay.io/cgwalters/coreos-assembler)


Note #119 should make this unnecessary.

Is coreos-assembler deprecated at this point as well?

cgwalters · 2018-06-19T14:30:30Z

I'm fine merging this as is, it doesn't need to be perfect, and will clearly change over time.

ashcrow · 2018-06-19T14:41:00Z

@cgwalters I agree it doesn't have to be perfect. However, I'm going to turn these commands into make targets, and update this doc. At that point I'd like to have it re-reviewed for merging.

ashcrow · 2018-06-19T15:18:27Z

/hold

ashcrow · 2018-06-19T17:34:41Z

Updated with make targets. I don't think it's quite ready to merge but close. Questions:

Is there a default for OSTREE_INSTALL_URL?
What about INSTALLER_TREE_URL?
Is this needed?

@lucab @jlebon @cgwalters @miabbott PTAL

ashcrow · 2018-06-19T17:34:49Z

/hold cancel

miabbott · 2018-06-19T18:39:31Z

@ashcrow OSTREE_INSTALL_URL and INSTALLER_TREE_URL are sourced from the pipeline-utils.groovy file.

See here: https://github.com/openshift/os/blob/master/pipeline-utils.groovy#L52

jlebon

This is cool, though it's missing the "linking" between each target. E.g. ideally make rpmostree-compose would pick up the RPMs we just created with make rdgo, and similarly, make os-image would point to the OSTree repo.

I think that's something we could do later though. At least we have the basic targets that codify what each step entails.

jlebon · 2018-06-19T18:32:22Z

Makefile

@@ -3,6 +3,10 @@ COMPOSEFLAGS ?=
 ROOT_DIR := $(shell dirname $(realpath $(lastword $(MAKEFILE_LIST))))
 CACHE_ARGS := $(shell if test -d cache; then echo $(shell pwd)/cache; fi)

+REPO ?= /tmp/repo


How about something like /srv/rhcos/repo?

Sure. Will update.

jlebon · 2018-06-19T18:34:56Z

README-development.md

- - ``export IMAGE=`ls /var/lib/imagefactory/storage/*.body```
- - ``qemu-img convert -f raw -O qcow2 $IMAGE rhcos-devel.qcow2``
+- Build packages from source repos: ``make rdgo``
+- Make the ostree: ``OSTREE_INSTALL_URL=<OSTREE_INSTALL_URL_HERE> make rpmostree-compose``


Minor: I think the more standard way of setting vars when calling make is make <target> VAR1=val1 VAR2=val2....

jlebon · 2018-06-19T18:41:36Z

INSTALLER_TREE_URL will use the Fedora installer tree by default, so it doesn't need to be subbed out (though it'd be much faster if pointed at a local mirror). OSTREE_INSTALL_URL... I guess we can hardcode for now the URL to the internal repo?

miabbott · 2018-06-19T18:43:08Z

The ostree init on line 59 will silently no-op if the repo is already initialized.

ashcrow · 2018-06-19T19:48:04Z

@miabbott Are you thinking that we should add that directory to the clean target to ensure it does not exist?

miabbott · 2018-06-19T19:57:32Z

@ashcrow It would probably make sense in terms of keeping things tidy.

ashcrow · 2018-06-19T20:34:32Z

@miabbott @jlebon updated

cgwalters · 2018-06-20T16:47:01Z

Makefile

+.PHONY: os-image
+os-image:
+	ostree --repo=repo remote add rhcos --no-gpg-verify ${OSTREE_INSTALL_URL}
+	ostree --repo=repo pull --mirror --commit-metadata-only rhcos


Hmm...I think if you're doing local builds you want to be using the local ostree repo. That's what I do.

The annoying pain point is it requires setting up a webserver. See projectatomic/rpm-ostree-toolbox#68 and there's other PRs/issues around this too.

cgwalters · 2018-06-20T16:52:32Z

A high level question here too is how much of this flow should live in https://github.com/cgwalters/coreos-assembler

I think having an opinionated container makes sense (versus delivering as an RPM as rpm-ostree-toolbox did).

Anyways...I'm OK to merge this as is too. Trying to change the Jenkinsfiles is going to be a whole other world.

cgwalters · 2018-06-20T16:54:17Z

README-development.md

- - ``export IMAGE=`ls /var/lib/imagefactory/storage/*.body```
- - ``qemu-img convert -f raw -O qcow2 $IMAGE rhcos-devel.qcow2``
+- Build packages from source repos: ``make rdgo``
+- Make the ostree: ``make rpmostree-compose OSTREE_INSTALL_URL=<OSTREE_INSTALL_URL_HERE> ``


I think this INSTALL_URL should be for make os-image right?

To be clear I'm OK to /lgtm after this small tweak, but up to you how much else you want to change.

Signed-off-by: Steve Milner <smilner@redhat.com>

ashcrow · 2018-06-21T16:26:05Z

Updated

cgwalters · 2018-06-21T16:42:03Z

/lgtm

lucab · 2018-07-09T11:26:15Z

I fear this has never been actually manually tried. There are a few gaps here and there (rpm-gpg assets, no valid OSTREE_INSTALL_URL default, mis-anchored paths in Makefile) which means that zero-knowledge onboarding here is still not a reality.

ashcrow · 2018-07-09T13:11:35Z

@lucab you are correct. There is work that needs to happen before someone who didn't actively work on the Jenkinsfiles and work on the underlying infrastructure can follow and recreate what's going on. There is also a side thing @cgwalters is doing in terms of changing the pipeline which may change some things.

Unfortunately I don't think we will be able to really finish #108 until the pipeline relaxes in flux once again.

ashcrow added help wanted Denotes an issue that needs help from a contributor. Must meet "help wanted" guidelines. kind/documentation Categorizes issue or PR as related to documentation. labels Jun 15, 2018

ashcrow requested review from lucab and jlebon June 15, 2018 13:52

openshift-ci-robot added the size/M Denotes a PR that changes 30-99 lines, ignoring generated files. label Jun 15, 2018

jlebon reviewed Jun 15, 2018

View reviewed changes

ashcrow mentioned this pull request Jun 15, 2018

Building locally is super confusing #108

Closed

cgwalters reviewed Jun 15, 2018

View reviewed changes

jlebon reviewed Jun 15, 2018

View reviewed changes

lucab reviewed Jun 18, 2018

View reviewed changes

README-development.md Outdated

- git

- coreos-assembler

- rpmdistro-gitoverlay

Copy link

Contributor

lucab Jun 18, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This additionally requires fedpkg and mock.

ashcrow force-pushed the build-docs branch from 71e4b32 to 6c21172 Compare June 18, 2018 13:50

ashcrow force-pushed the build-docs branch 3 times, most recently from 963931a to 77b7e5e Compare June 18, 2018 14:54

miabbott reviewed Jun 18, 2018

View reviewed changes

jlebon mentioned this pull request Jun 18, 2018

WIP: Allow rdgo logic to run local or in CI #114

Closed

cgwalters reviewed Jun 19, 2018

View reviewed changes

openshift-ci-robot added the do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. label Jun 19, 2018

ashcrow force-pushed the build-docs branch from 77b7e5e to 5cd2803 Compare June 19, 2018 17:31

openshift-ci-robot removed the do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. label Jun 19, 2018

ashcrow changed the title ~~README-development.md: Initial how to build info~~ Initial how to build info and Make targets Jun 19, 2018

jlebon reviewed Jun 19, 2018

View reviewed changes

ashcrow force-pushed the build-docs branch from 5cd2803 to 93b92e0 Compare June 19, 2018 19:51

ashcrow force-pushed the build-docs branch from 93b92e0 to f62a62b Compare June 19, 2018 20:34

cgwalters reviewed Jun 20, 2018

View reviewed changes

ashcrow added 2 commits June 21, 2018 09:37

README-development.md: Initial how to build info

add5b05

Makefile: Move developer into make targets

0c36718

Signed-off-by: Steve Milner <smilner@redhat.com>

ashcrow force-pushed the build-docs branch from f62a62b to 0c36718 Compare June 21, 2018 13:38

openshift-ci-robot assigned cgwalters Jun 21, 2018

openshift-ci-robot added the lgtm Indicates that a PR is ready to be merged. label Jun 21, 2018

openshift-merge-robot merged commit d55750f into openshift:master Jun 21, 2018

Initial how to build info and Make targets #107

Initial how to build info and Make targets #107

Conversation

ashcrow commented Jun 15, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ashcrow commented Jun 18, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

cgwalters commented Jun 19, 2018

ashcrow commented Jun 19, 2018

ashcrow commented Jun 19, 2018

ashcrow commented Jun 19, 2018 • edited Loading

ashcrow commented Jun 19, 2018

miabbott commented Jun 19, 2018

jlebon left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jlebon commented Jun 19, 2018

miabbott commented Jun 19, 2018

ashcrow commented Jun 19, 2018

miabbott commented Jun 19, 2018

ashcrow commented Jun 19, 2018

Choose a reason for hiding this comment

cgwalters commented Jun 20, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ashcrow commented Jun 21, 2018

cgwalters commented Jun 21, 2018

lucab commented Jul 9, 2018

ashcrow commented Jul 9, 2018

ashcrow commented Jun 15, 2018 •

edited

Loading

ashcrow commented Jun 19, 2018 •

edited

Loading