Installing rabbitmq without using the distro version is failing and it works if
you use the distro version. So let's standardize on that for both ubuntu and
RHEL.
Change-Id: If757cf57574bb8389ae270523bfc3772d8dc08db
Signed-off-by: Lance Albertson <lance@osuosl.org>
- since we are using distro_packages for rabbitmq the default rabbitmq
config name is broken overriding this attribute fixes this
Change-Id: Ia732d5ac4c64eab2126bbc385998585c96ff23d9
There is a bug in the rabbitmq-server setup for Queens that may cause
the cluster to lock up and stop responding[0]. The workaround for this
issue is to disable the management plugin, so add an option that allows
us to deploy this workaround.
[0] https://bugs.launchpad.net/bugs/1783203
Change-Id: I1d0e6c5fb8eeb35e4e966ef102e46c45c357a1ca
Add attribute to allow mq cluster to be built on-the-fly rather
than searching nodes.
Change-Id: I1196f95465c0cb21824954feb261b9d1cfd84d26
Closes-Bug: #1412886
The problem that this change addresses is that the address_for method
will not work correctly if there are multiple IP address associated
with the specified interface.
The approach to solving this problem and moving towards the overall
goal of having one place where service networking information is
stored is to convert address_for calls into endpoints, and add a
address() method to the endpoints interface for IP address resolution.
The address() method has the following behavior: if the
bind_interface of an endpoint is set, then the IP is looked up on
the interface. Otherwise, the IP specified in the host attribute is
returned. This allows the caller to choose either method of
determining what IP a service will be bound to.
This initial change switches both the openstack-ops-database and
openstack-ops-messaging cookbooks over to use endpoints instead of
address_for. The other cookbooks will be switched over time.
blueprint increase-ip-binding-flexibility
Change-Id: I4a610409b9542a4c802f94b557299bb97dd0781b
Fixes: bug 1282041
Implements: blueprint centos-rhel-for-ops-messaging
On CentOS/RHEL, the OS base repos don't have a package for
rabbitmq-server. The rabbitmq cookbook by default expects to be able to
install the latest rabbitmq-server rpm from the rabbitmq site
(use_distro_version = false). However, in our cookbooks we are overriding
'use_distro_version' to true, which works for ubuntu as it has a
rabbitmq-server package in it's base repo
This commit basically leaves the use_distro_version flag as false unless
ubuntu/suse is the platform_family, so that CentOS/RHEL can grab the rpm
from the rabbitmq site
Change-Id: I4e31cbb4957f0e696a1528fa3f7c6c80695989ac
This change removes the attributes we'll be moving to openstack-common
and changes reference of those attributes to reflect new (more
consistent) namespace.
Change-Id: I1602d2e6241a62b188d64fa760786fb95a734404
Implements: blueprint clean-up-attr-for-mq-and-db
The node[openstack][*][rabbit] attributes for all the services using rabbitmq
(block-storage, compute, image, metering, network) should default to the same
values as node[openstack][mq]. Bumping the major version accordingly.
Change-Id: Id20a48a3ef2506622239e8e6f15ce1f76060b828
Since this breaks people using rabbit in the standard way, I
rather revert this change, and fork my own ops-messaging cookbook
since this is intended as an example cookbook (granted others
may use it directly). This code would be better off in my
own fork, where i do not impact others, with a non-merged upstream
rabbitmq feature.
This reverts commit 3b35ac204d.
Change-Id: I53da70583c50ae985d3842c3d709eb81f565d63d
This change integrates against a rabbitmq cookbook feature which
is not yet merged upstream. However, given the default code path
does not change, felt okay to get this out there. Especially b/c
we are dependant on it. This feature allows one to run rabbit
and clustering across a specific address. For more details, see:
http://tickets.opscode.com/browse/COOK-3320
Change-Id: I2a10c551cb2176dadaee8842eeadfdfc8cb93e44
Now that `openstack-common` handles the `["openstack"]["mq"]`
attributes, there is no need to set them in the default test cases.
Change-Id: I1987a6c4ae0898bcf2cd0100fd1f5408990f2592
Clustering is set to false by default. When enabled, will setup
rabbitmq disk node clustering. Also, added opscode to attribution,
somehow had removed them in the previous rebase.
Change-Id: Idd249d0d950f5f4eb44cd7c10566ece254ed7f45
We had two reviews which added an ops-messaging cookbook. Updating
this cookbook with patterns we estabished from the database cookbook.
Craig already did most of the work, however, updated the tests to
use the chefspec find_provider, added clustering support, and attempted
to follow the service_type attribute patterns used in database cookbook.
Since these cookbooks are not complete, added a larger commit than
would have.
Change-Id: I0edfc996df413c6e41a31d53f964d7f5fc9e0aeb
This initial commit provides support for an openstack-ops-messaging
cookbook. In this initial release, we provide the capability to install
and configure RabbitMQ. This cookbook is modeled after the
openstack-ops-database cookbook in the way in which it can be utilized.
Included are tests for all recipes and this release supports Ubuntu
only.
Change-Id: I4d799fa46e67e0f2eca2ac8b205f51c07e0df01e
Ghanshyam Mann