Jump to: navigation, search

XenServer/XenServer CI/AdminTips

Status

The status report of the CI is available here: http://f0ab4dc4366795c303a0-8fd069087bab3f263c7f9ddd524fce42.r22.cf1.rackcdn.com/ci_status/results.html This report is updated by a cron job, which is part of osci (see below).

Components

  • infrastructure.hg This repo contains the high level installation scripts. http://hg.uk.xensource.com/openstack/infrastructure.hg
  • install-nodepool Contains the low-level installation scripts, used by infrastructure.hg/osci/*
  • nodepool Responsible for pre-baking nodes
  • project-config Scripts for baking the nodes
  • openstack-citrix-ci Services to run tests on the nodes
  • openstack-xenapi-testing-xva Scripts to build the ubuntu appliance
  • xenapi-os-testing Main entry point for the tests
  • devstack-gate Used to run the tests

Description

Nodepool service will launch an instance in the Rackspace cloud, and run the XenServer related scripts from project-config/nodepool/scripts. The responsibilities of those scripts are:

  • Convert the instance to a xenserver
  • Install a virtual appliance inside the xenserver. The virtual appliance is created by a service running inside Citrix. The appliance is created with the scripts living at openstack-xenapi-testing-xva and the produced appliances are stored here: http://downloads.vmd.citrix.com/OpenStack/xenapi-in-the-cloud-appliances/ and in RackSpace containers as well.
  • prepare the node for running OpenStack tests - this is "standard OpenStack magic"
  • once preparation finished, shut down the node

After nodepool ran those scripts, it will take a snapshot of that node, so that further instances could be launched quicker. Unfortunately the upstream nodepool is not prepared for this lifecycle, so we are running a custom version of nodepool.

The second service is citrix-ci, which will get nodes from the pool and run https://github.com/stackforge/xenapi-os-testing/blob/master/run_tests.sh on them. This script will customise the cached version of devstack-gate.

The third service is citrix-ci-gerritwatch that is listening for gerrit events and communicates with citrix-ci.

Main box

This is the instance which orchestrates the execution of tests.

  • Authentication via ssh keys. For access, join the XenAPI team meeting on IRC.
  • All the components are deployed to this instance.
  • created with scripts living inside infrastructure.hg/osci (this is a citrix -internal repository)
  • IP address is also stored within the repository

openstack-citrix-ci

  • source code: https://github.com/citrix-openstack/openstack-citrix-ci
  • /opt/osci/src is where sources live
  • /opt/osci/env is the python virtual environment where it's installed
  • /etc/osci/osci.config is the configuration file
  • /var/log/osci/citrix-ci-gerritwatch.log is the log file for gerritwatch
  • /var/log/osci/citrix-ci.log is the log file for osci
  • /var/log/osci/status_upload.log is the log file for the status-upload cron job
  • A separate user, osci is used to run all the services. If you want to access the python environment, do the following:
sudo -u osci -i
. /opt/osci/env/bin/activate
  • For update, use the scripts provided at infrastructure.hg/osci.

service: citrix-ci-gerritwatch

This service watches the gerrit stream and adds jobs to the queue

  • Logs: /var/log/osci/citrix-ci-gerritwatch.log
  • (Re)start: (re)start the service citrix-ci-gerritwatch

service: citrix-ci

This service progresses jobs through the lifecycle (see below)

  • Logs: /var/log/osci/citrix-ci.log
  • (Re)start: (re)start the service citrix-ci
  • three threads:
    • main (run through state machine, see below)
    • collect results (long blocking process, so its split out so main stuff makes progress -- service net?)
    • delete node (go to collected, delete is a pain, tries up to to 10 mins, need to fix the delete process!!!!!)

cron job: upload status

  • Uploads the CI status to swift
  • Can be disabled by touching the file /etc/osci/skip_status_update
  • see:
crontab -l -u osci

Utilities

These utilities are available if you activated the environment. Some of the utilities are available without activating the environment by using symlinks/wrapper scripts. These are: osci-manage and osci-view.

  • osci-check-connection - should be removed.
    • checks to node, to xenserver, then logs uploaded to swift from both
  • osci-manage
    • what citrix-ci service runs
    • runs main lifecycle
    • can also manually add a job to the queue
  • osci-upload
    • upload logs from local host directory up to swift
    • called from osci-manage (via python not cli)
  • osci-watch-gerrit
    • what citrix-ci-gerrit-watch runs
    • reads gerrit stream and adds jobs to DB
  • osci-create-dbschema
    • creates job DB
  • osci-run-tests - could be used to run the tests on a node.
    • print out what commands would be executed, use the "print" executor:
      osci-run-tests [print|exec] user host ref
    • To run the tests, use a exec executor:
      osci-run-tests exec jenkins 162.242.171.81 refs/changes/86/131386/1 openstack/nova
    • calls out to host to run tests
    • used (via python not cli) by osci-manage
  • osci-view
    • prints out the job DB in useful ways

To report status:

  • osci-view is called and content uploaded to swift
  • /opt/osci/src/openstack-citrix-ci/upload-ci-status

nodepool

service: nodepool

Provisions VMs to use in the tests

  • Logs: /var/log/nodepool/nodepool.log, /var/log/nodepool/debug.log
  • (Re)start: killall nodepool; rm /var/run/nodepool/nodepool.pid; start nodepool

Utilities

to get information/control the nodepool service.

  • To list the images:
nodepool image-list

To see how its all doing:

nodepool list

To claim a node:

nodepool hold id

project-config

These scripts are used by nodepool to prepare the nodes. Please see that nodepool's configuration refers to the location of these scripts.

Useful commands

  • osci-view list: Gives current queue, what is running etc. Shouldn't have jobs in here that are 'older' than 2 hours unless they are 'Finished'.
  • nodepool list: Gives a list of the currently available nodes. Should have some nodes that are 'Ready' or 'Building'
  • eval `ssh-agent`; ssh-add ~/.ssh/citrix_gerrit; osci-manage -c 12345/1; ssh-agent -k: Queue job 12345, patchset 1

VM lifecycle

  • Queued -> Running: citrix-ci job has got a new node from nodepool (nodepool list will show it as 'held') does osci-run-tests to hand over to xenapi-os-testing
  • Running -> Collecting: Job has finished; citrix-ci has changed state to Collecting - waiting on log collection thread
  • Collecting -> Collected: Log collection thread has posted logs to swift and updated job with logs URL
  • Collected -> Finished: Citrix-ci has posted to gerrit and the job is now complete
  • <anything> -> Obsolete: a new job for the same change (recheck or new patchset) has been added

Code


TODO list

  • start using stackforge http://git.openstack.org/cgit/stackforge/xenapi-os-testing/
  • reduce number of excluded tests, using above stackforge integration
  • stop forking infra projects, where possible
  • consider moving to zuul, talk to turbo hipster folks
  • create new environment to test out config changes
  • create system to construct dev environments
  • trial out neutron based tests (with quark?)
  • BobBall to send johnthetubaguy an example on how to deploy the Main Box