| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
* Add db-setup to the 'Afters' for ceilometer
* Improve order of units in 'Afters'
* Remove deprecated syslog.target
* Add missing Wants=network-online.target
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The OpenStack services write their log to stderr even when use_syslog
was set to True in the configuration file. This is then captured by
journald, resulting in duplicated entries on the log. Set
StandardOutput=null and StandardError=null in the unit files to prevent
this.
As the services log is being dumped on journald, there is no need to
duplicate the log in separate log files, so remove --log-file from the
unit files as well.
Last but not least, enable writing to the syslog for Ironic and
Keystone.
|
| |
|
|
|
|
|
|
| |
And fix some existing as well. Generate a separate postgres.conf file
so that we don't have to reference keystone on the postgres setup
systemd unit and Ansible script.
|
| |
|
|
|
|
|
|
| |
This way it is possible to avoid running the setup script in subsequent
boots, either by disabling the setup service using systemctl, or
removing the respective configuration file from /etc/openstack/.
|
|
|
|
| |
Also ensure that the /etc/openstack directory exists.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Those variables were already defined somewhere in the cluster, so there
is no need to duplicate them.
|
|
|
|
| |
To be consistent with the other IP addresses used.
|
| |
|
| |
|
|
|
|
| |
Change-Id: I5000f08e91b19f26ce3d4265cd7c08be9098bcfc
|
|
|
|
|
|
|
| |
When the server is configured to listen connections only from some
IP addresses, the unit fails if the network is not configured and
there aren't interfaces with this IP.
Change-Id: Ibc4a00f5cc8efbbd8edd2df227272b9420438c6c
|
|
|
|
|
|
|
|
| |
Its not needed anymore.
This completes commit
eb2c33af1b56b322524d058dd6f560cb21d98a0d
Change-Id: Icfc73bb2fd3cf3140561b2192a59ab6ec13e266b
|
|
|
|
|
|
|
|
|
|
|
| |
Some interesting changes:
- The command hwclock(8) uses the drift correction for all relevant
operations and it is unnecessary yo specify --adjust on the hwclock
command line now
- zramctl(8):
this NEW COMMAND allows to control /dev/zramN devices
Change-Id: I196d3a9a0f38bcb10c2adf10a200acc4d4218a9b
|
|
|
|
| |
Change-Id: Ibc954e855452d83fe364083455bc2d5616457cbb
|
|
|
|
|
|
|
|
| |
Cinder needs to have enabled the api service in the controller node
in order to work. This means also run the config-setup and the
db-setup services for it.
Change-Id: I5e79ebf8ce75ef7320b3177b953966095a0884ae
|
|
|
|
| |
Change-Id: I065dcd8926775c5469830af25c3fc6a49d89e432
|
|
|
|
|
|
| |
So we have all the xcb libraries together
Change-Id: I31712c4a68d1b75a104bcf758989f63b3aca39a0
|
|
|
|
|
|
| |
This includes a fix for vulnerability CVE-2014-9495
Change-Id: Ia62f889ad532b8614c4ec4536f517c5eb303416d
|
|
|
|
|
|
| |
it isn't built with SSL/https support by default
Change-Id: I5c124da046aea2802193790a51d94bf1f75e490b
|
|
|
|
|
|
|
|
|
|
| |
Depending on openstack-keystone is wrong, as the admin user, admin
tenant, admin role, service tenant and endpoint, are only created after
starting and enabling the openstack-keystone service. Not having those
created will cause keystone commands executed by some Ansible scripts to
fail.
Change-Id: I54288f3de2814dbb6639a894d55d7395f052a428
|
|
|
|
|
|
|
|
|
| |
Otherwise postgres-server and keystone-setup services will initially
fail, as keystone-setup needs a postgres-server running to succeed, but
the postgres-server can only run after some configuration, which was
previously being done by keystone-setup.
Change-Id: I2d649d494cb54119e3b9bd3d9f6deb46bfb2dd12
|
|
|
|
| |
Change-Id: I27f2e61a411174541bf2e20a1fd5c30ca8f8ff4e
|
|
|
|
|
|
|
|
|
| |
The OpenStack Ironic Conductor service requires that an iSCSI initiator
is running on the same local machine. This commit configures and sets
an iSCSI initiator to run on the controller node on a two-node and
three-node OpenStack system. For a one-node system, this service was
already configured as it is required by OpenStack Cinder.
Change-Id: I46291d9cd1a31e3ff91888401d45ab0dc67e3677
|
|
|
|
|
|
|
| |
Also avoid running it on the compute machine on a two-node OpenStack
system.
Change-Id: I091bb4641291aaf1d699be2bd433cbda1f87d743
|
|
|
|
| |
Change-Id: I3ab72d36b6b164088e8d31ac4e2359377714a9fd
|
| |
|
|
|
|
| |
Reviewed-By: Sam Thursfield <sam.thursfield@codethink.co.uk>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
fe15d7e Calculate the build graph in worker-build
3d87709 Make distbuild use an ArtifactReference not an Artifact
internally when
5239f51 Don't serialise the entire build graph
aa6dfcb Use --verbose/-v to show build output, --debug/-d for morph
debug
124e2f1 distbuild: Add test suite for distbuild-helper
8103617 distbuild: Don't check if there is free disk space on the
initiator
9d27007 Add distbuild-trove-nfsboot.write
006096c distbuild: Don't create a directory for build output until we
get some
52ff3a3 distbuild: Kill the whole process tree when cancelling a build
b2f13f7 deploy: Clean up `morph deploy` documentation a bit
f163d9c deploy: Deploy and upgrade systems from the same 'cluster'
definition
1c01f18 deploy: Factor out a bit of code into its own function
f323af1 distbuild: Move SubprocessEventSource into its own module
40c9378 distbuild: Fix log message when listening for connections
Change-Id: I97c38b65db62b0631745ac5d47d572d0218fb9ff
|
|
|
|
|
|
|
|
| |
Connecting to the nfs server would fail because one of the services
would segfault, because nfs-utils used a pointer before checking whether
it was NULL, and gcc would later optimise away the null pointer check.
Change-Id: Ia007fc9c33012daf56f65677ab138f706a7a0263
|
|
|
|
|
|
|
| |
Other systems can use overlayfs, but we can't run Linux 3.18 on Highbank
so must use unionfs-fuse for now instead.
Change-Id: I26a4c1656298e071df31684bfe1dd616ecbf4558
|
|
|
|
| |
Change-Id: I828aa856fc14b04d3e4d358f78deb0d8abc22229
|
|
|
|
|
|
|
|
|
|
| |
This will be used by future versions of Morph.
By adding OSTree to the cross-bootstrap systems we have significantly
increased the size and complexity of them. Some of this can be reduced:
OStree doesn't actually depend on all of 'foundation', just 'glib'.
Change-Id: I89403bf4625178e6f887402b5817f6a727cfcf97
|
|
|
|
| |
Change-Id: I14177e899e404fdc72728f694f8551d5f8518f4b
|
|
|
|
|
|
| |
Also make morph depend on PyGObject, it will be needed for OSTree.
Change-Id: Icfa9abb95f884ca9b1dd720648567bd704e74d85
|
|
|
|
| |
Change-Id: Iec8bb1d659b037b45b8733f46d3f4af2413631a7
|
|
|
|
| |
Change-Id: I2555a2ee2c6b3b7e4243b985343bc8afb2432c77
|