Download code has been refactored. That allow now take exact
commit over TOOLS_COMMIT for testing new version.
ADD_CENT_REPO and UPDATE_FUEL order has been changed
to reflect real order of that operations.
MOS versions <=7.0 should prevent `stop deployment` operation when
environment has already deployed. This requires that clusters have
`deployed_before` atribute in generated attributes of environment.
This script sets deployed_before in True value for all environments.
This script should be run on master node.
As starting from 9.0 the access to the Fuel node with SSH is
restricted to the PXE/Admin network only, it is necessary to
add an access rule after the Fuel node bootstrap. This will
let everyone to access Fuel from everywhere. Also updated
the jenkins job configuration.
Added functionality to make role-based and particular nodes updates,
i.e. update only computes or ceph-osds. This is needed to divide
update process into stages which is usefull on very big environments.
This could be also useful to rerun the update process on a particular
node. Also the time interval variable is added. Using this variable
it is possible to update some senstive part of a cluster giving a
time to synchronize its states.
sys.stdout.write() does not guarantee that the output will show up,
as buffers need to be explicitly flushed. The more suitable solution is
to use the print_function from __future__.
Fixed the following errors:
* SSHD is inaccessible in 8.0 as it is only configured to listen ADMIN_NETWORK
* Mode HA is no longer an option
* Changed output of some of functions to more informative one
As this job is used in jenkins as non-stop job
(i.e. it should be running until aborted), there is a possibility
that executor thread will receive more than one interrupt,
and will generate SIGTERM during the cleanup procedure.
This commit adds some signal ignorance to the cleanup
procedure to have a chance of successful completion if
multiple signals are arriving.
Jenkins' job build_cluster is using no database for a while,
as everything is being obtained from libvirt directly.
So this commit gets rid of database usage along with removal
of unnecessary checks. Also fixed grammar and some PEP8 issues.
New job configuration is introduced to get effective build managment.
Also, changed the order of boot devices for virtual machines, as
sometimes there might be a problem with re-bootstraping slaves that were
provisioned with OS previously.
As we hardcode "peerdns=no" in eth0 during master node deployment
a dirty hack is needed to make resolver work. This change introduces new
option "DNS1=$ADMIN_SUBNET_GATEWAY" to ifcfg-eth1, although this address
is not in eth1-network it will work anyway.
* Changed structure to classful
* Got rid of custom packages installation
* Support for 5.1.1, 6.0, 6.1, 7.0 added
* Added logging to file
* Better wording in help message
* Changed way how multiple repos are handled in apt-get
* 5.1.1 now uses dist-upgrade as it's necessary for
the installation of nova and qemu.
Added new options:
- ability to guess fuel version and master ip
- ability to install more than one repo
- ability whether to download repos automatically
New repositorie will be used exclusive, e.g. updated packages
will be searched only in selected repositories.
This is initial commit, which consists of md5checker.py tool and
md5checker.dat database. Database has checksums from Fuel 6.0,
6.0-mu3, 5.1.1 and 5.1 releases and will be replenished in future.
As murano uses config injection in settings.py we should
reinject murano settings after openstack-dashboard update.
This patch-set introduces murano check/fix section in update
script.
Small unrelated changes also have been made:
Now script on each node uses `logger` with tag mos_apply_mu to
log messages. This means that output of script from each node
will apear on master-node in
/var/log/remote/$nodename$/mos_apply_mu.log
Also amount of output was reduced. Messages became clearer.
Master IP is a mandatory option now.
As for now this script can make VMs, Volumes and Networks
in libvirt. Also it manages subnets in order to achieve
zero crossing between them. Fuel master node gets dynamic
network parameters from script by injection during bootloader
stage. Public network being injected in Fuel-master node,
so it can be accessible from network. Cluster's network
is configured with apropriate values.
Included changes in nailgun network & settings configuration.
Now script generates bash-script, sends it to remote node and executes it.
New options has been added. Totaly renewed help/usage.
Flake8 runs w/o errors.
Code Review / tools / sustaining.git / log