There have been problems with using uWSGI for lava-server, including
preventing updates of 2016.6-2~bpo8+1 in jessie-backports.
The LAVA software team have completed a change which will be included
in the 2016.11 release which removes the need for the
libapache2-mod-uwsgi, libapache2-mod-wsgi and uwsgi-core packages. The
replacement is the gunicorn package.
This does involve changes to the apache configuration of all sites
which upgrade to 2016.11.
Ahead of installing 2016.11, two apache modules need to be enabled.
https://staging.validation.linaro.org/static/docs/v2/installing_on_debian.h…
$ sudo a2enmod proxy
$ sudo a2enmod proxy_http
$ sudo service apache2 restart
These will not affect older installations.
Then install 2016.11 and your instance will continue to operate.
If you have made changes to
/etc/apache2/sites-available/lava-server.conf then the changes to the
upstream file will need to be merged into your changes.
If you choose to remove libapache2-mod-uwsgi after completing the
install, be aware that there is a Debian bug which may affect you.
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=839666
If your apache server fails to restart after removing the
libapache2-mod-uwsgi package, you will need to workaround this bug by
executing:
$ sudo a2dismod uwsgi
$ sudo apache2ctl restart
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
2016.9 includes a rewrite of each of the daemons used by LAVA,
lava-server, lava-master and lava-slave. When installing the upgrade,
admins need to accept the maintainer's version of the scripts in
/etc/init.d/ for these daemons. Where previously changes were made in
the scripts in /etc/init.d/, the same parameters can now be changed in
associated configuration files to prevent these conflicts in the
future.
/etc/init.d/lava-slave - change /etc/lava-dispatcher/lava-slave
/etc/init.d/lava-master - change /etc/lava-server/lava-master
/etc/init.d/lava-server - change /etc/default/lava-server
The prompt would look something like this:
Configuration file '/etc/init.d/lava-server'
==> Modified (by you or by a script) since installation.
==> Package distributor has shipped an updated version.
What would you like to do about it ? Your options are:
Y or I : install the package maintainer's version
N or O : keep your currently-installed version
D : show the differences between the versions
Z : start a shell to examine the situation
The default action is to keep your current version.
*** lava-server (Y/I/N/O/D/Z) [default=N] ? y
See https://validation.linaro.org/static/docs/v2/pipeline-server.html#configuri…
References in the changelogs, commit history and announcements which
start with LAVA- followed by a number relate to the planning stories
which the LAVA team use for new features. The LAVA planning has now
been opened up to anonymous access by appending the reference to a
stub URL: http://projects.linaro.org/browse/, e.g.
http://projects.linaro.org/browse/LAVA-62
The planning stories do not support comments or changes other than by
the LAVA software team. If you have comments on the stories, please
use the lava-users mailing list.
lava-server changes
Show measurements and units of testcase detail page
Drop lxc schema - no support for alternatives
Configure logging for linaro_django_xmlpc calls
Fix link to lava-tool context-help
Fix the "Debian-based distribution" link grammar
Improve monitor test type, allow multiple tests
LAVA-494 allow searching for device tags in tables
Add support for LXC with multinode
Fix broken ref link for unit_test
Fix NoReverseMatch from ImageReports2.0 editing.
Fix typo in ./share/validate.py --hostname
Update hikey jinja2 templates for LXC
Fix occasionally failing multinode tests.
Add a configuration file for lava-master
Update lava-slave and lava-master docs
LAVA-374 - drop versiontools
Tidy up some build errors and reformat consistently.
Improve documentation describing inline test definitions
Major improvements for the "Writing MultiNode" page
Add docs for the lava-vland-names change
Using BOOTX in Juno bootcmds
Correcting default uboot commands in juno device template
Document the code locations for developers
Ensure multinode description is generated correctly
Allowing overrides to connection and action timeouts
master: Fix log rotations
Remove filtering by filter id for available testcases in image reports.
Init: remove the right lock file and simplify
Adding uboot jinja2 templates for all Juno flavours
LAVA-739 - declare sub_id and job.id for multinode
master: don't validate jobs on the master
select_device only accepts pipeline jobs
Remove legacy SyntaxHighlighter from job definition pages.
lava-master: simplify log file handling
Fix title disappearance on image reports.
Ignore retired devices in validate call
Don't add anchors to logs for running jobs
Don't crash when log contains invalid test names
Schema: accept 'parse' for inline test definitions
Use libYAML when loading logs
LAVA queries use cases.
Bug link fix for url regex.
Add LogEntry for BugLinks.
Update the init scripts
Remove the entry point as it's no longer needed
Ensure test-case name is valid
Extend vland unit tests to cover assignment
Expand on the device dictionary and templates
Expand the notes on using jessie-backports
Add a Contents page
Rewrite lava-server command line from scratch
events: set umask to a restrictive value
Bug links port.
LAVA-200 - track admin actions made in the UI
Add device tags support to V2
Allow the monitor test type
Update qemu command line options for arm64
lava-dispatcher changes
Catch the edge case where serial line corruption means we miss the testrun
Add support for alternative Test actions
Update manpage of lava-slave for configuration change.
Allow configuring the slave from file
git: use -C instead of --git-dir
Allow setting the envrionment on most systems
Handle exceptions from subprocess from compression
download: don't expand the path when retrying
Grab finalize from the root pipeline
Log exception as strings and not objects
LAVA-740 - export vlan names and interface names
Reopen log file when rotating the logs
LAVA-699 - Support hi6220-hikey board with lxc
Init: remove the right lock file and simplify
pipeline:vland: raise JobError if switch_id or port_id is None
Remove redundant assignation
Allow logging at the validate stage
Force the shell to produce an initial prompt
Handle errors if the testrun start is omitted.
Pipeline parser: sort and reverse in one call
logs: remove unused support for local logs
slave: put all tmp files into the same directory
Power: improve logging message
Exclude patterns from description
Port lava-slave to Python3
ApplyOverlayTftp: remove duplicated untar_file
ExportDeviceEnvironment: only source valid shell_file's
add LEDE to available distributions
Always log the command to be run beforehand.
There is currently a delay on uploading lava-server to
jessie-backports due to problems with uwsgi in Debian. A backport will
be made available once these problems are fixed. In the meantime, the
same build as will later appear in jessie-backports is available in
the production-repo:
https://validation.linaro.org/static/docs/v2/installing_on_debian.html#lava…https://tracker.debian.org/pkg/lava-serverhttps://tracker.debian.org/pkg/lava-dispatcherhttps://qa.debian.org/developer.php?email=pkg-linaro-lava-devel%40lists.ali…
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
A number of the configuration files which manage the daemons used by
lava-server and the lava-slave daemon for lava-dispatcher are being
updated in 2016.9. This has implications for those who install the
upgrade, especially if this is normally done by automation.
Configuration file '/etc/init.d/lava-slave'
==> Modified (by you or by a script) since installation.
==> Package distributor has shipped an updated version.
What would you like to do about it ? Your options are:
Y or I : install the package maintainer's version
N or O : keep your currently-installed version
D : show the differences between the versions
Z : start a shell to examine the situation
The default action is to keep your current version.
*** lava-slave (Y/I/N/O/D/Z) [default=N] ? y
Installing new version of config file /etc/init.d/lava-slave ...
dpkg will check the checksums of the init files against the current
and previous packages and halt installation with the above prompt in
case of doubt. Admins may need to install the update manually
(directly on the command line).
The recommendation is to accept the package maintainer's version but
all admins may want to check the difference between the two.
Changes in this release include fixing the log rotate handling of
lava-master using reload and a rewrite of the lava-server CLI script
to drop the use of entrypoints which changes the syntax of the daemon
options.
The most common reason for init files to be modified is to change the
loglevel. This can be done in /etc/default/lava-server:
# Additional options that are passed to the Daemon.
DAEMON_OPTS="loglevel=debug"
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
LAVA has long been developed in the open as far as the git
repositories and bug reports are concerned. However, the git commit
history can often include references to LAVA-NNN and these references
have previously been obscure.
Anonymous access has now been granted to all LAVA-* references, so
here's some context on what and where.
https://projects.linaro.org/projects/LAVA
This is a JIRA instance and we refer to issues in JIRA as cards or stories.
All stories in LAVA are now visible to the community, so you can map
any of the LAVA-NNN references directly to the actual URL by
prepending projects.linaro.org/browse/LAVA-NNN, e.g.
https://projects.linaro.org/browse/LAVA-734
It has not been possible to open the actual organisation of the
stories (KANBAN) and some elements do not provide the same links as
when viewed internally but hopefully some of those issues can be
addressed in the future.
For now, community members can follow any explicit LAVA-NNN reference
from the git history or elsewhere to the actual story and view the
complete description and all comments on that story. Most stories also
provide a direct link to the review which implemented that story
within the Comments section. Merged reviews provide information in the
"Included In" menu about which branches and tags include that commit.
Tags match the production release of the same name as the tag, so this
allows everyone to know which releases contain which changes.
Stories which are currently status Open may simply be ideas which
still need to be scoped. Some of these will be dropped without being
implemented as use cases and code development moves on.
Dates in stories are very rough estimates, usually longer than it will
actually require.
If community members have comments or questions on particular stories,
please use the lava-users mailing list.
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
This release moves to requiring django 1.8 from jessie-backports.
(django1.7 is no longer supported, django1.8 is the LTS release)
See https://lists.linaro.org/pipermail/lava-announce/2016-June/000010.html
As part of this release, OpenID and therefore Launchpad login support
has been removed - the openid support for django cannot support
django1.8.
See https://lists.linaro.org/pipermail/lava-announce/2016-June/000011.html
This release also includes support for running lava-server with the
recently released django1.10 which will soon be available in Debian
Stretch (current testing).
2016.8 includes a rewrite or large sections of the documentation and
the new logging support for LAVA V2.
2016.8 includes the changes planned for 2016.7, so this is a larger
set of changes than previous releases.
lava-server changes
lava-master: use also get_env_string for multinode
Handle unrecognised result messages.
Publisher: drop privileges at startup
Add lava-publisher init scripts
Update mustang jinja template
Similar jobs feature.
Django1.10 fixes
v2: include a 'secrets' field in the job def
Adding device-type templates for juno
Fix a deprecation warning with render_to_string
events: add more details and use a useful username
Ensure failed health checks go directly to offline.
Faster loading of yaml logs
Add a u-boot-commands timeout just for panda
Improve error handling in result metadata
use job.id inside a not job.is_multinode conditional
Fix multinode link from definition back to the job.
Allow parentheses in test case names
Allow the d02 debian installer grub device to be overridden
in device-dictionary
Allow for creating devices already offline.
Device state transition validation.
Open context-sensitive help in a new browser tab
Unavailable qemu command should not fail unit tests.
LAVA-719 - support branding of source and bugs URL
Ensure logging to django logs is info or higher
Fix e1d66f to use pk when not multinode.
Create and display measurements with units
result: don't crash when parsing an invalid result
Implement notification blacklist.
results: handle skip result
Simple notification list.
Add 'name' to testcase export.
Use the right syntax for character delays
First device configuration for ST b2120h410
Fix bug #2278 - inconsistent multinode job id / alias usage
Fix HTTP500 by allowing for + in test case names
Show job sub_id for multinode jobs.
Implement IRC notifications.
Fix bug #2263 - parameters and params reference in job def and test def
LAVA-708 - Device path should be a list
Fix a crash when viewing a query for the first time
result: show the metadata as a list (and sublists)
Rename conflicting notification properties.
log: don't show 'extra' result data
log: add a link to each line using AnchorJS
log: skip broken strings
log: add an icon for the download button
log: add link from the result page back to the log
Results: improve admin page
TestResults: order by job_ids then name
Simplify a bit the result page
Remove unnecessary loading of django-tables
log: redirect complete_log to the job_detail page
log: add a link to the result page for each result
log: fix HTML syntax errors
log: improve rendering of errors and exceptions
Update load addresses for larger multiplatform kernels
Fix result table
log: fix a bug when the page is reloaded
Protect from admin error in health check submission
Fix default value for device_path to be None and not 'None'.
Add missing device_path to nexus jinja templates.
log: adapt the result parser to the new log stream
mustang UBoot needs 32bit header
logs: update job status and device information
log: change the arrow when clicking on the affix
Fix handling of context with multinode
Fix metadata handling for multinode and dynamic connections
Fix hidden-device-type listings in JobTable
job: remove redudant information
Events: add a monitoring thread
Initial notifications for v2.
lava-master: save the logs in output.yaml
job: add a new template for the new log format
LAVA-262 Allow admins to expire user accounts
log: better formatting of tracebacks
Remove support for Django < 1.8
Improve scheduler debug with device details.
.
Documentation updates
Add links and notes to developer branch guide
Add notes on making Lava Test Shell portable
Add notes on running lava-server unit tests
Add timeout documentation.
Update the developer guide
Document the 'secrets' dictionary
Ensure V2 documentation examples are available.
update local user account image
tidy up api docs
Remove multinode use cases
tidy up the writing-multinode page
expand simple-admin for admin roles
tidy up hidden toctree listings for previous/next markup
Update chapters for theme
Switch to the bootstrap theme
updates for multinode and simple administration
Major update to the docs for writing multinode tests
move all examples into one directory and add test definitions
move lava tool issues to a separate file
fold the FAQ into the lava-tool docs
update the multinode use cases
port the mustang example to a separate yaml file
use rst macros for see also
Add publishing API ref doc
initial content for a results intro
Move doc yaml to a directory
WIP rewrite of the multinode doc
Start thinking about how to grow a lab
Re-org some early admin stuff
Split out the completed YAML jobs
Query omit documentation updates.
Fix documentation for test definition name handling
add instruction for -t jessie-backports
move example YAML to an rtsi for easier checking
add notes on setting up the first device and device type
fix whitespace in migration example
Update the scheduling ordering with links
Add notes on LAVA being developer focused
Update other examples for deploy change
fixup deploy action
add example of first qemu V2 device
start the pipeline design page
Minor wording tweaks
Rework the hacking session doc
expand notes on first installation
tweaks and updates for writing tests
Fix definition link to log for pipeline
Updates for test repositories
update multinode docs for V2
fix build messages and errors
update examples of params support and custom scripts for parsing
complete fixme in advanced-installation
add background on CI and LAVA
add notes and images for first job submission and results
explain the first job and tidy up the example YAML
Clean up health check docs
add notes for first job
Significant cleanup of wording around lava-test-shell
Add lots of code-block:: yaml directives
Add details of features and architecture.
Add content to the what-is section
lava-dispatcher changes
Only require conversion parameters if using conversion.
Support some basic kernel conversion tasks
Export the 'secrets' dictionary to the overlay
uboot: only log self.errors when it's not empty
Enabling NFS only deployment
Fix connection_timeout handling for lava-test-shell
validate needs to set errors, not raise
Fixes for multinode without needing ACK
Device environment: fix a crash if env_dut is None
Unavailable qemu command should not fail unit tests.
Use character_delays instead of character-delays
Tests: Use the name defined in the job definition
Show when a revision is being applied for VCS
Fix bug #2263 - parameters and params reference in job def and test def
Improve logging and use lazy logging
Log more test results metadata
LAVA-708 - Device path should be a list
Support kickstart installations on grub, and centos distro type
Allow for a prefix with nfsrootfs
Finalize: send the job status with the right level
log: send the full exception untouched
beaglebone-black: update load addresses for larger multiplatform kernels
Take into account, device_path that is unset.
Fix YAML formatting of wait string log
change the result logging format
Add sample nexus9 device dictionary.
LAVA-701 - Support vendor flashing using fastboot.
utils/messages.py: fix out of index error
Add support for fixup dictionary and patterns
Ensure guest command is available in multinode
Stop V2 test shells needing an ACK from the dispatcher
logging: add datetime and a better structure
Upgrade the failure test case for lavabot
First batch of VExpress changes
Handle ValueError while waiting
To enable jessie-backports, simply copy your existing apt source for
jessie and change jessie to jessie-backports, then run apt update.
Installations from backports are not automatic, you'll need to tell
apt to select jessie-backports for django1.8, as outlined in previous
mails to this list:
$ sudo apt install -t jessie-backports python-django
python-django-common python-django-tables2
Once 2016.8 migrates into Stretch and is then backported to
jessie-backports itself, apt will be able to handle the extra
dependencies.
https://tracker.debian.org/pkg/lava-serverhttps://tracker.debian.org/pkg/lava-dispatcherhttps://qa.debian.org/developer.php?email=pkg-linaro-lava-devel%40lists.ali…
The production repo also has this release.
https://validation.linaro.org/static/docs/v2/installing_on_debian.html#lava…https://validation.linaro.org/static/docs/v1/installing_on_debian.html#lava…
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
Due to resource shortages in the team, conference time and a few
problematic fixes which still need testing, we will have to skip the
delayed 2016.7 release and move all changes to the 2016.8 release, due
at the end of the first week of August 2016.
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
OpenID has continued to be available for lava-server on jessie as long
as lava-server stayed on django1.7 but the OpenID django support does
not operate with django later than 1.7. To continue with upstream
django security and bug fix support, lava-server needs to move to
requiring django1.8 which means that OpenID can no longer be supported
in jessie. Support for OpenID in unstable and testing was removed when
django1.8 support first arrived.
Once lava-server 2016.7 arrives in jessie-backports, it will conflict
with the python-django-auth-openid package - this means that to
install the 2016.7 upgrade from jessie-backports, apt will first cause
the removal of python-django-auth-openid. (Allowing
python-django-auth-openid to remain on the system will corrupt
subsequent django operations, causing the package installation to fail
during the database migrations.)
OpenID support in lava-server is also due to be removed for 2016.7, so
the documentation and settings will no longer reference OpenID.
LDAP support, Debian SSO and local django accounts will remain available.
https://packages.debian.org/stable/python-django-auth-openidhttps://wiki.debian.org/DebianSingleSignOn
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
Further to the notice about the availability of django1.8 support:
https://lists.linaro.org/pipermail/lava-announce/2016-April/000007.html
The 2016.7 release of lava-server will be moving to requiring
django1.8 from jessie-backports and will no longer work with django1.7
available in jessie.
This has been done because the 1.7 tree is no longer receiving
upstream security or bug fix support.
lava-server installations newer than 2016.2 can use django1.8 already,
as long as python-django-tables2 is updated at the same time. So you
are invited to upgrade to django1.8 at any time prior to upgrading to
2016.7.
To add jessie-backports, copy the apt source for your favourite Debian
mirror and paste that as a new line with jessie updated to
jessie-backports.
e.g. for the http://mirror.bytemark.co.uk/debian mirror:
deb http://mirror.bytemark.co.uk/debian jessie main
deb-src http://mirror.bytemark.co.uk/debian jessie main
becomes
deb http://mirror.bytemark.co.uk/debian jessie main
deb http://mirror.bytemark.co.uk/debian jessie-backports main
deb-src http://mirror.bytemark.co.uk/debian jessie main
$ sudo apt update
$ sudo apt install -t jessie-backports python-django python-django-tables2
$ sudo apache2ctl restart
$ sudo service lava-server restart
It is imperative that python-django-tables2 is updated at the same
time as python-django - the old version of django will not work with
new tables and the old version of tables will not work with new
django.
Backports do not automatically replace packages you already have
installed, so you remain in control of which backports get installed
onto your systems and can run apt dist-upgrade without getting extra
upgrades.
Backports related to lava-server are also fully supported and are the
recommended (soon to be required) way of installing updates for
lava-server.
After 2016.7, updates of lava-server will not install on jessie
without django and django-tables being updated from jessie-backports.
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
lava-server changes
* Update V1 docs for Ubuntu changes - lava-server no longer
migrates into Ubuntu and was removed from Xenial.
* Drop heartbeat support
* Prevent scheduler ValueError in reservation
* scheduler: reduce the number of SQL queries
* Expose DISALLOWED_USER_AGENTS to handle search bots
* Add a page for listing Pipeline Devices.
* Add Auth support in REST API for more functions
* Remove device status glyphicons everywhere, since heartbeat is dead.
* Create metadata on the number of test definitions
* Remove the need for extensions
* Remove deprecated lava_projects
* Update docs for guestfs and resulting issues.
* Enable job definition metadata.
* dispatcher-master: support zmq CURVE encryption
* Add documentation on using ZMQ curve
lava-dispatcher changes
* Load the overlay as an extra QEMU drive to prevent
need for loopback devices.
* Support delays between sending characters for all actions
* LXC support for fastboot and adb and drop the need for adb or
fastboot to be installed.
* Remove adb related code
* Port V2 unit tests only to python3
* ZMQ: Support encryption of the logs sent to the server
* Add support for QEMU Debian Installer tests
* Update standards version (no changes)
* Remove adb and fastboot from recommends and add python-guestfs as a
dependency.
To use LXC with LAVA 2016.6 (V2 only), the lxc package itself needs to
be installed from jessie-backports (the lava-dispatcher package has a
version requirement in the Recommends which enforces this, if lxc is
installed). Additionally, for the new version of LXC to work
correctly, the dispatcher itself needs to be rebooted to apply the LXC
updates to the kernel boot arguments.
To enable jessie-backports, simply copy your existing apt source for
jessie and change jessie to jessie-backports, then run apt update.
Installations from backports are not automatic, you'll need to tell
apt to select jessie-backports for the LXC package:
apt install -t jessie-backports lxc
https://validation.linaro.org/ has already been updated to 2016.6.
Packages have been uploaded to Debian (unstable) and will appear on
the mirrors in due course. Once these versions migrate into stretch,
backports will be made to jessie-backports.
https://tracker.debian.org/pkg/lava-serverhttps://tracker.debian.org/pkg/lava-dispatcherhttps://qa.debian.org/developer.php?email=pkg-linaro-lava-devel%40lists.ali…
The production repo also has this release.
https://validation.linaro.org/static/docs/v2/installing_on_debian.html#lava…https://validation.linaro.org/static/docs/v1/installing_on_debian.html#lava…
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
The 2016.6 release which is currently in preparation adds some
important dependencies which some admins may prefer to install ahead
of the upgrade.
Debian has the concept of Dependencies which must be installed and
Recommends which are optional but expected to be useful by most users
of the package in question. The Recommends for GuestFS can be omitted
from the installation if admins desire but this needs to be done ahead
of the upgrade to 2016.6.
V2 in 2016.6 adds support for GuestFS which removes the need for
loopback support for QEMU devices. GuestFS supports a wide range of
filesystem utilities and other virtualisation support tools, not all
of which will be of interest to LAVA. (Notably, libguestfs Recommends
a mail-transport-agent and the default MTA in Debian is exim. With the
default exim4 config, this simply adds a local MTA with no external
support.)
Debian supports opting out of Recommends when installing packages, so
if admins have concerns about extra packages being installed on the
dispatchers (e.g. if using ARMv7 dispatchers or simply to reduce the
complexity of the install) then Recommends can be omitted for the
installation of these dependencies:
The option is:
$ sudo apt --no-install-recommends install python-guestfs
We'll add a note to the installation docs for this too.
With the move to V2 having a dumb dispatcher, the dependencies
themselves do need to be installed to allow the server to dictate what
jobs the dispatcher can run without the dispatcher failing due to
missing support. If later LAVA upgrades need any of the packages that
would have been installed as Recommends, LAVA will add an explicit
dependency on just those packages and any others will remain optional.
--
Neil Williams
=============
neil.williams(a)linaro.org
http://www.linux.codehelp.co.uk/
