The Kolla design is for services to use the internalURL for
service to service communication. In Mitaka, Neutron added
a new config parameter specifying which URL to use to to
contact Nova, making the default 'public'. This patch sets
the value to 'internal'.
Change-Id: I2d36f3b4a860af9e9034ebfb2b5cea56450e5e4e
Closes-Bug: #1565624
At high scale, such as 64 nodes with 13TB ram and 2600 cores, nova
seems to struggle when scheduling 100+ VMs at the same time. The
issue is unrelated to the database, as the error printed indicates
the max_scheduling_attempts have been reached. Increase that value
to something more fitting of a 100 node cluster.
Change-Id: I8982d77c7c66db8f7c95b9fd73f58ceb66dbd723
Closes-Bug: #1563664
Previously the code looked at mariadb.pid, but this seemed
flakey in function. It seemed racey and prone to failure on
slower connections to a registry. The original task was
extremely complex and it didnt really verify that the MariaDB
was ready to serve connections. Use wait_for with a regex
instead.
Change-Id: I3aafac04f03639b08e0ef4d6a9c9e1a4499f000c
Closes-Bug: #1564278
This patch set makes "kolla-ansible prechecks" flag an error if
any password is empty in /etc/kolla/passwords.yml.
Change-Id: I87dee25b79c97be64ca49a5638c7f5a30d4cf464
Closes-Bug: #1563506
Added general_log to ansible/roles/mariadb/templates/galera.cnf.j2
to improve mariadb logging.
This will be helpful to debug mariadb issues especially when
mariadb is scaled.
Test results of this patch set are at:
http://paste.openstack.org/show/492852/
Change-Id: I80438d1bbdd1ed2a1f47489c6f9c45b8107340a0
Closes-Bug: #1563668
Currently Heka writes the keepalived logs in
/var/log/kolla/haproxy/keepalived.log.
This commit changes this to /var/log/kolla/keepalived/keepalived.log.
Closes-Bug: #1565499
Change-Id: I3033097bd77ddbf72948697b34a6a499ea903083
Add a nova-ssh container to handle the `nova migrate` and
`nova resize` case, in which the nova will use ssh to copy
files between machines.
Change-Id: Ie6675943f3aeabfbba8589d308d55b9c89d732db
Closes-Bug: #1562141
To be kolla deploy multiple clouds, we need to be able to configure
virtual_router_id other wise haproxy will fail setup the VIP for the
second cloud.
Partially-Implements: blueprint multiple-cloud
Closes-Bug: #1564547
Change-Id: I9eb27dd6fba61205841eadafc96601e235d2fe6d
Currently the delegate_to doesnt happen and the neutron role creation is
attempted once on the first server and is skipped. The re-ordering of hosts in
site.yml seems to make the first host to be one inside neutron-server group
yielding the expected results. This patch needs to be re-visited as soon as a
version of ansible is chosen that fixes the issues with delegate_to
Co-Authored-By: Steven Dake <stdake@cisco.com>
Co-Authored-By: Vikram Hosakote <vhosakot@cisco.com>
Co-Authored-By: Nate Potter <nathaniel.potter@intel.com>
Co-Authored-By: Ganesh Mahalingam <ganesh.mahalingam@intel.com>
Change-Id: Ia712b323aa9d750d470a11ee899ab1b3054a903f
Partial-Bug: #1546789
When a node uses two physical interfaces for its two VIPs, these
physical interfaces should be tied together, so both VIPs will
be taken out of scheduling if either one fails. Without this change,
if a request comes into one interface that needs access to the
second interface to process the request, the original request
unnecessarily fails. Repeating this results in a black hole where
a failing server keeps getting new requests.
Change-Id: Ic51e6584c1fbda3eb7821cb47f759c77e562cc65
Closes-Bug: #1550455
On AIO installation we cannot assume that the public IP address
will be the first entry in "getent ahostsv4" result, because
it may be also a localhost address. To make this check positive
in AIO, we should look for the public IP in the whole output.
Change-Id: I1da7b95d7f00c7f87ff68ead46bf55fdea812599
Closes-Bug: 1564564
Scale limit testing on a 64 node cluster with 13 TB RAM and 2600
cores showed that atleast 1800 database connections are required
to appropriately start 2000 virtual machines simulateounsly. Other
documentation on the internet recommends larger values such as 8000,
so we set to a larger value that should be able to handle a maxed out
512GB ram per compute node cluster with all services enabled.
Change-Id: I8767cf3fb04e066cc22e796c647e944b4e4a1742
Closes-Bug: #1564275
haproxy-system-wrapper is a solution for systemd from upstream. it can
handle the reload graceful.
Change-Id: I6a3d141af065e429bd1be1b7252f5c6df1fda3bb
Closes-Bug: #1559238
Checking the fix condition first in the when clause rather than the
dynamic condition. Otherwise it will raise issue, especially in
multinode env.
For example use
when:
- nova_console == 'novnc'
- inventory_hostname in groups['nova-novncproxy']
rather than
when:
- inventory_hostname in groups['nova-novncproxy']
- nova_console == 'novnc'
Because the nova-novncproxy groups may do not exist.
Change-Id: I1e76ee239908a17a4c2ffd9a18c570fbc485172e
Closes-Bug: #1563643
Kolla recommends LTS wily kernel version 4.2 on Ubuntu 14.04.
http://docs.openstack.org/developer/kolla/quickstart.html#installing-dependencies
The precheck check the running kernel version, because checking
the linux-image-generic-lts-wily cannot be a valid check.
The package (linux-image-generic-lts-wily) can be installed, but
the running kernel can be other version.
Closes-Bug: #1563500
Co-Authored-By: Vikram Hosakote <vhosakot@cisco.com>
Change-Id: I4e08bca95493b2dbd0ad5f8d1ea4e07b18c4959d
