This tool will contain more things eventually but for
now it'll just cater for a list of things to clean up
in environments with low disk space.
Change-Id: I84568f4de20c6c38c96e7a4c1fcad6d5c949d205
If you have a cloud where some of your stacks are stuck in
DELETE_FAILED state and you need to run something to clean
up all the resources, this tool is designed for this.
At the moment, it just handles nested stacks but it can be
extended to handle more things (i.e. network ports, etc). It
does come with destructive behaviour so the operator needs
to be careful using it.
Change-Id: I5c419aa0f45d243b585c9794406086891bfe06c1
This change adds support for the SHELVED_OFFLOADED and (partly)
VERIFY_RESIZE states. Shelved-offloaded instances don't have a
hypervisor associated with them. VERIFY_RESIZE instances should still
be running on their associated hypervisor. They are also present, but
"shut off", on another hypervisor. Even with this change, the code
still generates a warning about the instance being present on multiple
hypervisors. Personally I find the warning quite useful, and I don't
really know what to do about it. Maybe we should suppress the warning
if the instance hasn't been in this state for long.
Change-Id: I3a56500bd350a070c96f719e702a5334ec8c8558
Cold stop: remove all network to agent associations
Cold start: add network to agent associations
Cold restart: do stop and start
Change-Id: I2494f6a1603d89a048afb697931dd84197af88e9
We found that sometimes in large deployments some
DHCP agents can be overloaded when others are almost
empty (especially if random scheduler is used for
DHCP agents).
That can cause issue with agents config syncronization
e.g. after agents restart.
This script can move some networks from overloaded agents
to less loaded and also remove network from some DHCP agents
if it is hosted on too many agents in same time.
Change-Id: Ib9ed1d75100ee66e04143eff4d30ccab1eb72abf
The trystack environment file sources credentials from OS_* env
variables and as such we need to pass them properly in tox.
Change-Id: I8abc7c4d4109a178a6547a87aec2c2dbff56b988
Add three new app_env options:
wp_latest
wp_cli
wp_importer
these can be used to override the default install locations
for wordpress and various utilities used by the playbooks,
ensuring that the playbook can be configured for use in
environments where network egress to the internet is limited.
Change-Id: I9db0d8e562fdb74c9c68cb9881d6a2bbb7e4e77a
Adds a very basic one-by-one livemigration tool to remove instances from
a host, but not quite as aggressively as the openstack CLI tool.
Change-Id: I54b03e7f7eb6eb0f4efc10de97dc208498f2a56b
[kubeadm](http://kubernetes.io/docs/getting-started-guides/kubeadm/) provides
an easier way to install kubernetes. Let's use it and simplify our terraform.
* swap out coreos for ubuntu/centos as kubeadm has apt/yum repos
* remove SSL creation as kubeadm does that
* remove etcd/cloudinit as kubeadm handles etcd
* add basic install script to set up deps and repos
tested installing to both ubuntu xenial and centos 7 images.
Change-Id: I63c2f0419457587c53b112c28ab96f55add7676a
A few folks reported that installing scsitools was throwning an error
for them because the apt cache on the system hadn't been initialized.
This patch ensures that we update the apt cache before attempting to
install scsitools.
The patch also makes the workload run faster overall: previously,
we updated the apt cache each time we installed a package. In this
patch, we break out the cache update into a separate task that runs
before any software is installed, thus eliminating several
time-consuming calls to "apt-get update".
Finally, this patch also makes the workload slightly more portable:
it replaces apt calls with package calls. Adding a new guest OS
family should just require adding a new cache update block (if
necessary for that OS) with an appropriate when statement.
Change-Id: I30ffbf2da13f22475f920328d84e64460834bcdb
Currently, the lampstack workload assumes that either you're using the
nova metadata service or that config drive is enabled globally in
your nova.conf file: there's no option to enable it on a per-environment
basis. This patch adds the ability to specify whether config_drive
should be used or not in the app_env variables, with a default value
set to "no" (which means that the new variable isn't required and that
if a user doesn't set it, he or she will get the same behavior as before
this patch).
Change-Id: I717bf5a37f98da79141af70df90feb7fd7b3f8c7
Creates a new tox env called profile in order for profiling not to be
mandatory for each workload run.
Change-Id: I699551f1d50776499ad67440bc0a6b71bd8101ef