Tarantool development patches archive
 help / color / mirror / Atom feed
From: Alexander Turenko <alexander.turenko@tarantool.org>
To: "Alexander V. Tikhonov" <avtikhon@tarantool.org>
Cc: Oleg Piskunov <o.piskunov@tarantool.org>,
	tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH v7] gitlab-ci: implement packing into MCS S3
Date: Thu, 30 Jan 2020 18:49:46 +0300	[thread overview]
Message-ID: <20200130154946.57toungzdlrfup5q@tkn_work_nb> (raw)
In-Reply-To: <b28804d4ab39b741dbc09f5c96c0670e16ef4c66.1580101688.git.avtikhon@tarantool.org>

CCed Oleg.

Oleg, please look at the commit message: whether it is understandable
w/o much context? If not, please, work together with Alexander on it.

I almost didn't look into the script (because I guess I'll no give much
value here and because Igor already did).

I have several small comments, but I don't against the patch at whole.

WBR, Alexander Turenko.

> The changes introduce new Gitlab-CI rules for creating packages on
> branches with "-full-ci" suffix and their subsequent deployment to the
> 'live' repository for master and release branches. Packages for tagged
> commits are also delivered to the corresponding 'release' repository.
> 
> The PackageCloud storage is replaced with the new self-hosted one

Packagecloud rules are not removed (and this is what we planned to do at
the moment), so the word 'replaced' looks confusing here.

> (based on S3 object storage) where all old packages have been synced.
> The new builds will be pushed only to S3 based repos. Benefits of the

This is not so, we don't plan to disable packagecloud repositories right
now.

> introduced approach are the following:
> * As far as all contents of self-hosted repos are fully controlled
> theirs layout is the same as the ones provided by the corresponding
> distro
> * Repo metadata rebuild is excess considering the known repo layout
> * Old packages are not pruned since they do not affect the repo
> metadata rebuilding time
> 
> For these purposes the standalone script for pushing DEB and RPM
> packages to self-hosted repositories is introduced. The script
> implements the following flow:
> * creates new metafiles for the new packages
> * copies new packages to S3 storage
> * fetches relevant metafiles from the repo
> * merges the new metadata with the fetched one
> * pushes the updated metadata to S3 storage

Is some blocking mechanism implemented? What will going on if two CI
jobs will fetch metadata, update them locally and them push back? Is it
depends on whether those CI jobs are run on the same machine or
different ones?

> 
> There are distro dependent parts in the script:
> * For RPM packages it updates metadata separately per each repo
> considering 'createrepo' util behaviour
> * For DEB packages it updates metadata simultaniously for all repos
> considering 'reprepro' util behaviour

What 'repo' means here? We have 1.10 repo. But it contains centos/7,
debian/jessie and other repos. I don't understand this paragraph, to be
honest.

Aside of that, we always update only one repository (say, 1.10 x
centos/7) at the moment. 'Simultaneous update' looks even more
confusing.

> 
> Closes #3380
> 
> @TarantoolBot
> Title: Update download instructions on the website
> 
> Need to update download instructions on the website, due to the new
> repository based on MCS S3.

First, it is stub, not a request to the documentation team. What actions
you are requested here?

Second, since we decided to keep current repositories structure
(separate 1.10/2.1/... repos and no version in a package name) no
actions are required actually.

> ---
> 
> Github: https://github.com/tarantool/tarantool/tree/avtikhon/gh-3380-push-packages-s3-full-ci
> Issue: https://github.com/tarantool/tarantool/issues/3380
> 
> v6: https://lists.tarantool.org/pipermail/tarantool-patches/2020-January/013763.html
> v5: https://lists.tarantool.org/pipermail/tarantool-patches/2020-January/013636.html
> v4: https://lists.tarantool.org/pipermail/tarantool-patches/2020-January/013568.html
> v3: https://lists.tarantool.org/pipermail/tarantool-patches/2019-December/013060.html
> v2: https://lists.tarantool.org/pipermail/tarantool-patches/2019-November/012352.html
> v1: https://lists.tarantool.org/pipermail/tarantool-patches/2019-October/012021.html

> diff --git a/.gitlab.mk b/.gitlab.mk
> index 48a92e518..243f83f2c 100644
> --- a/.gitlab.mk
> +++ b/.gitlab.mk
> @@ -98,14 +98,38 @@ vms_test_%:
>  vms_shutdown:
>  	VBoxManage controlvm ${VMS_NAME} poweroff
>  
> -# ########################
> -# Build RPM / Deb packages
> -# ########################
> +# ###########################
> +# Sources tarballs & packages
> +# ###########################

But it is about packages, not tarballs. It seems you copied the comment
from .travis.yml, but it is not appropriate here.

> +
> +# Push alpha and beta versions to <major>x bucket (say, 2x),
> +# stable to <major>.<minor> bucket (say, 2.2).
> +GIT_DESCRIBE=$(shell git describe HEAD)
> +MAJOR_VERSION=$(word 1,$(subst ., ,$(GIT_DESCRIBE)))
> +MINOR_VERSION=$(word 2,$(subst ., ,$(GIT_DESCRIBE)))
> +BUCKET=$(MAJOR_VERSION)_$(MINOR_VERSION)
> +ifeq ($(MINOR_VERSION),0)
> +BUCKET=$(MAJOR_VERSION)x
> +endif
> +ifeq ($(MINOR_VERSION),1)
> +BUCKET=$(MAJOR_VERSION)x
> +endif

Let's push to 2.1, we'll add 2x for compatibility on using redirects on
download.tarantool.org.

>  
>  package: git_submodule_update
>  	git clone https://github.com/packpack/packpack.git packpack
>  	PACKPACK_EXTRA_DOCKER_RUN_PARAMS='--network=host' ./packpack/packpack
>  
> +deploy: package
> +	for key in ${GPG_SECRET_KEY} ${GPG_PUBLIC_KEY} ; do \
> +		echo $${key} | base64 -d | gpg --batch --import || true ; done

I guess we need just secret key to signing.

> +	./tools/update_repo.sh -o=${OS} -d=${DIST} \
> +		-b="s3://tarantool_repo/live/${BUCKET}" build

I would hide name of an S3 bucket under a CI variable. It is more about
our infrastructure and it would be good to show less in the repo.

> +	for tag in $$(git tag) ; do \
> +			git describe --long $${tag} ; \
> +		done | grep "^$$(git describe --long)$$" >/dev/null && \

Simpler way:

 | git name-rev --name-only --tags --no-undefined HEAD 2>/dev/null

See https://stackoverflow.com/a/11489642/1598057

> +		./tools/update_repo.sh -o=${OS} -d=${DIST} \
> +			-b="s3://tarantool_repo/release/${BUCKET}" build

> +function get_os_dists {
> +    os=$1
> +    alldists=
> +
> +    if [ "$os" == "ubuntu" ]; then
> +        alldists='trusty xenial bionic cosmic disco eoan'
> +    elif [ "$os" == "debian" ]; then
> +        alldists='jessie stretch buster'
> +    elif [ "$os" == "el" ]; then
> +        alldists='6 7 8'
> +    elif [ "$os" == "fedora" ]; then
> +        alldists='27 28 29 30 31'
> +    fi

We have no Fedora 27 in CI. Should it be here? I don't mind, just found
the tiny inconsistency.

> +
> +    echo "$alldists"
> +}
> +
> +function prepare_ws {
> +    # temporary lock the publication to the repository
> +    ws_suffix=$1
> +    ws=${ws_prefix}_${ws_suffix}
> +    ws_lockfile=${ws}.lock
> +    if [ -f $ws_lockfile ]; then
> +        old_proc=$(cat $ws_lockfile)
> +    fi
> +    lockfile -l 60 $ws_lockfile
> +    chmod u+w $ws_lockfile && echo $$ >$ws_lockfile && chmod u-w $ws_lockfile
> +    if [ "$old_proc" != ""  -a "$old_proc" != "0" ]; then
> +        kill -9 $old_proc >/dev/null || true
> +    fi

So the current script can be killed by another instance of the script?
This means that the lock does not work.

> +
> +    # create temporary workspace with repository copy
> +    $rm_dir $ws
> +    $mk_dir $ws

Am I understand right: we don't create a copy of remote repository?

  parent reply	other threads:[~2020-01-30 15:49 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-27  5:13 [Tarantool-patches] [PATCH v6] " Alexander V. Tikhonov
2020-01-28 13:18 ` [Tarantool-patches] [PATCH v7] " Igor Munkin
2020-01-30 15:49 ` Alexander Turenko [this message]
2020-01-31  4:59   ` Alexander Tikhonov
2020-01-31 22:53     ` Alexander Turenko
2020-02-01 18:55       ` Alexander Tikhonov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200130154946.57toungzdlrfup5q@tkn_work_nb \
    --to=alexander.turenko@tarantool.org \
    --cc=avtikhon@tarantool.org \
    --cc=o.piskunov@tarantool.org \
    --cc=tarantool-patches@dev.tarantool.org \
    --subject='Re: [Tarantool-patches] [PATCH v7] gitlab-ci: implement packing into MCS S3' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox