Welcome to the Free Software contributions diary of Loïc Dachary. Although the posts look like blog entries, they really are technical reports about the work done during the day. They are meant to be used as a reference by co-developers and managers. Erasure Code Patents StreamScale.

gf-complete test coverage report

The make coverage target is added to gf-complete to create a lcov report while running the tests with make check.

The full report is archived.

Posted in jerasure | Leave a comment

Teuthology docker targets hack (4/4)

The teuthology container hack is completed by adding a flag to retrieve packages from a user specified repository instead of gitbuilder.ceph.com. The user can build packages from sources and run a job, which will implicitly save a docker image with the package installed. The second time the same job is run, it will go faster because it reuses the image. For instance the following job:

machine_type: container
os_type: ubuntu
os_version: "14.04"
suite_path: /home/loic/software/ceph/ceph-qa-suite
roles:
- - mon.a
  - osd.0
  - osd.1
  - client.0
overrides:
  install:
    ceph:
      branch: master
  ceph:
    wait-for-scrub: false
tasks:
- install:
    repository_url: http://172.17.42.1/trusty
- ceph:

runs under one minute:

{duration: 47.98, flavor: basic, owner: loic@dachary.org, success: true}
Posted in ceph, docker | 3 Comments

Building Ceph Debian GNU/Linux packages

The following script explains how to create Debian GNU/Linux packages for Ceph from a clone of the sources.

releasedir=/tmp/release
rm -fr releasedir
mkdir -p $releasedir
#
# remove all files not under git so they are not
# included in the distribution.
#
git clean -dxf
#
# git describe provides a version that is
# a) human readable
# b) is unique for each commit
# c) compares higher than any previous commit
# d) contains the short hash of the commit
#
vers=`git describe --match "v*" | sed s/^v//`
#
# creating the distribution tarbal requires some configure
# options (otherwise parts of the source tree will be left out).
#
./autogen.sh
./configure --with-rocksdb --with-ocf --with-rest-bench \
    --with-nss --with-debug --enable-cephfs-java \
    --with-lttng --with-babeltrace
#
# use distdir= to set the name of the top level directory of the
# tarbal to match the desired version
#
make distdir=ceph-$vers dist
#
# rename the tarbal to match debian conventions and extract it
#
mv ceph-$vers.tar.gz $releasedir/ceph_$vers.orig.tar.gz
tar -C $releasedir zxf ceph_$vers.orig.tar.gz
#
# copy the debian directory over and remove -dbg packages
# because they are large and take time to build
#
cp -a debian $releasedir/ceph-$vers/debian
cd $releasedir
perl -ni -e 'print if(!(/^Package: .*-dbg$/../^$/))' ceph-$vers/debian/control
perl -pi -e 's/--dbg-package.*//' ceph-$vers/debian/rules
#
# always set the debian version to 1 which is ok because the debian
# directory is included in the sources and the upstream version will
# change each time it is modified.
#
dvers="$vers-1"
#
# update the changelog to match the desired version
#
cd ceph-$vers
chvers=`head -1 debian/changelog | perl -ne 's/.*\(//; s/\).*//; print'`
if [ "$chvers" != "$dvers" ]; then
   DEBEMAIL="contact@ceph.com" dch -b -v "$dvers" "new version"
fi
#
# create the packages
# a) with ccache to speed things up when building repeatedly
# b) do not sign the packages
# c) use half of the available processors
#
PATH=/usr/lib/ccache:$PATH dpkg-buildpackage -j$(($(nproc) / 2)) -uc -us

Continue reading

Posted in ceph, debian | 2 Comments

jerasure.org installation notes

The jerasure.org is setup to host the upstream repositories for the GF-complete and jerasure libraries. Contributors may sign-up or re-use their existing GitHub account. A companion continous integration server runs make check on each merge request.
Continue reading

Posted in gitlab, jerasure | Leave a comment

Customizing the gitlab home page

The customization of the Gitlab home page is a proprietary extension that is not available in the Free Software version. When running Gitlab from docker containers, the home page template needs to be moved to a file that won’t go away with the container:

$ layouts=/home/git/gitlab/app/views/layouts/
$ docker exec gitlab mkdir -p /home/git/data/$layouts
$ docker exec gitlab mv $layouts/devise.html.haml /home/git/data/$layouts
$ docker exec gitlab ln -s /home/git/data/$layouts/devise.html.haml \
   $layouts/devise.html.haml

The template can now be modifed in /opt/gitlab/data/home/git/gitlab/app/views/layouts/ from the host running the container. It is a HAML template which can have raw HTML as long as proper indentation is respected.

Continue reading

Posted in gitlab | 1 Comment

A make check bot for Ceph contributors

The automated make check for Ceph bot runs on Ceph pull requests. It is still experimental and will not be triggered by all pull requests yet.

It does the following:

A use case for developers is:

  • write a patch and send a pull request
  • switch to another branch and work on another patch while the bot is running
  • if the bot reports failure, switch back to the original branch and repush a fix: the bot will notice the repush and run again

It also helps reviewers who can wait until the bot succeeds before looking at the patch closely.
Continue reading

Posted in ceph, docker | Leave a comment

Teuthology docker targets hack (3/4)

The teuthology container hack is improved so each Ceph command is run via docker exec -i which can read from stdin as of docker 1.4 released in December 2014.
It can run the following job

machine_type: container
os_type: ubuntu
os_version: "14.04"
suite_path: /home/loic/software/ceph/ceph-qa-suite
roles:
- - mon.a
  - osd.0
  - osd.1
  - client.0
overrides:
  install:
    ceph:
      branch: master
  ceph:
    wait-for-scrub: false
tasks:
- install:
- ceph:

under one minute, when repeated a second time and the bulk of the installation can be reused.

{duration: 50.01510691642761, flavor: basic,
  owner: loic@dachary.org, success: true}

Continue reading

Posted in ceph, docker | Leave a comment

Why are by-partuuid symlinks missing or outdated ?

The ceph-disk script manages Ceph devices and rely on the content of the /dev/disk/by-partuuid directory which is updated by udev rules. For instance:

  • a new partition is created with /sbin/sgdisk --largest-new=1 --change-name=1:ceph data --partition-guid=1:83c14a9b-0493-4ccf-83ff-e3e07adae202 --typecode=1:89c57f98-2fe5-4dc0-89c1-f3ad0ceff2be -- /dev/loop4
  • the kernel is notified of the change with partprobe or partx and fires a udev event
  • the udev daemon receives UDEV [249708.246769] add /devices/virtual/block/loop4/loop4p1 (block) and the /lib/udev/rules.d/60-persistent-storage.rules script creates the corresponding symlink.

Let say the partition table is removed later (with sudo sgdisk --zap-all --clear --mbrtogpt -- /dev/loop4 for instance) and the kernel is not notified with partprobe or partx. If the first partition is created again and the kernel is notified as above, it will fail to notice any difference and will not send a udev event. As a result /dev/disk/by-partuuid will contain a symlink that is outdated.
The problem can be fixed by manually removing the stale symlink from /dev/disk/by-partuuid, clearing the partition table and notifying the kernel again. The events sent to udev can be displayed with:

# udevadm monitor
...
KERNEL[250902.072077] change   /devices/virtual/block/loop4 (block)
UDEV  [250902.100779] change   /devices/virtual/block/loop4 (block)
KERNEL[250902.101235] remove   /devices/virtual/block/loop4/loop4p1 (block)
UDEV  [250902.101421] remove   /devices/virtual/block/loop4/loop4p1 (block)
...
Posted in ceph | Leave a comment

How many PGs in each OSD of a Ceph cluster ?

To display how many PGs in each OSD of a Ceph cluster:

$ ceph --format xml pg dump | \
   xmlstarlet sel -t -m "//pg_stats/pg_stat/acting" -v osd -n | \
   sort -n | uniq -c
    332 0
    312 1
    299 2
    326 3
    291 4
    295 5
    316 6
    311 7
    301 8
    313 9

Where xmlstarlet loops over each PG acting set ( -m “//pg_stats/pg_stat/acting” ) and displays the OSDs it contains (-v osd), one by line (-n). The first column is the number of PGs in which the OSD in the second column shows.
To restrict the display to the PGs belonging to a given pool:

ceph --format xml pg dump |  \
  xmlstarlet sel -t -m "//pg_stats/pg_stat[starts-with(pgid,'0.')]/acting" -v osd -n | \
  sort -n | uniq -c

Where 0. is the prefix of each PG that belongs to pool 0.

Posted in ceph | 3 Comments

Gitlab CI runner installation

The instructions to install GitLab CI runner are adapted to Ubuntu 14.04 to connect to GitLab CI and run jobs when a commit is pushed to a branch. The recommended packages are installed except postfix and with ruby2.0 and ruby2.0-dev in addition:

sudo apt-get update -y
sudo apt-get install -y wget curl gcc libxml2-dev libxslt-dev \
   libcurl4-openssl-dev libreadline6-dev libc6-dev \
   libssl-dev make build-essential zlib1g-dev openssh-server \
   git-core libyaml-dev libpq-dev libicu-dev \
   ruby2.0 ruby2.0-dev

Ruby2.0 is made the default ruby interpreter

sudo rm /usr/bin/ruby /usr/bin/gem /usr/bin/irb /usr/bin/rdoc /usr/bin/erb
sudo ln -s /usr/bin/ruby2.0 /usr/bin/ruby
sudo ln -s /usr/bin/gem2.0 /usr/bin/gem
sudo ln -s /usr/bin/irb2.0 /usr/bin/irb
sudo ln -s /usr/bin/rdoc2.0 /usr/bin/rdoc
sudo ln -s /usr/bin/erb2.0 /usr/bin/erb
sudo gem update --system
sudo gem pristine --all

The bundler gem is installed

sudo gem install bundler

and the GitLab CI runner user created

sudo adduser --disabled-login --gecos 'GitLab CI Runner' gitlab_ci_runner

The GitLab CI runner code is installed in the home of the corresponding user with:

sudo su gitlab_ci_runner
cd ~/
git clone https://gitlab.com/gitlab-org/gitlab-ci-runner.git
cd gitlab-ci-runner
bundle install --deployment

The CI token is retrieved from the GitLab CI pannel

and used to grant access to the runner:

CI_SERVER_URL=http://workbench.dachary.org:8080 \
  REGISTRATION_TOKEN=778b1d4856f26da392a bundle exec ./bin/setup

The daemon is started from root with:

su gitlab_ci_runner -c 'cd $HOME/gitlab-ci-runner ; bundle exec ./bin/runner'

The GitLab CI interface shows the runner as registered:

Assuming all the above was done from within a docker container, it can be persisted as an image with

docker commit b504ab6ba122 gitlab-runner

and used to multiply the runners with:

$ docker run --rm -t gitlab-runner \
  su gitlab_ci_runner -c  'cd $HOME/gitlab-ci-runner ; \
  CI_SERVER_URL=http://workbench.dachary.org:8080 \
  REGISTRATION_TOKEN=b14852619da392a \
  bundle exec ./bin/setup ; bundle exec ./bin/runner'
Registering runner with registration token: 2619da3, url: http://workbench.dachary.org:8080.
Runner token: 35f9d40f2e072487870f987
Runner registered successfully. Feel free to start it!
* Gitlab CI Runner started
* Waiting for builds
2014-12-06 17:18:26 +0000 | Checking for builds...nothing
2014-12-06 17:20:27 +0000 | Checking for builds...received
2014-12-06 17:20:27 +0000 | Starting new build 6...
2014-12-06 17:20:27 +0000 | Build 6 started.
2014-12-06 17:20:32 +0000 | Submitting build 6 to coordinator...ok
2014-12-06 17:20:33 +0000 | Completed build 6, success.
2014-12-06 17:20:33 +0000 | Submitting build 6 to coordinator...aborted
2014-12-06 17:20:38 +0000 | Checking for builds...nothing
...

When the container is stopped, the runner must be manually removed from the GitLab CI. Projects in the GitLab CI will be confused by the disapearance of the runner and must be removed and re-added otherwise no job will get scheduled.
It is easier to install on Fedora 20

sudo gem install bundler

sudo useradd -c 'GitLab CI Runner' gitlab_ci_runner
export PATH=/usr/local/bin:$PATH
cd ~/
git clone https://gitlab.com/gitlab-org/gitlab-ci-runner.git
cd gitlab-ci-runner
bundle install --deployment
CI_SERVER_URL=http://workbench.dachary.org:8080 \
  REGISTRATION_TOKEN=XXXXX bundle exec ./bin/setup
Posted in gitlab | Leave a comment