| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
| |
|
|
|
|
| |
We don't use UPSTREAM_TROVE_USER or UPSTREAM_TROVE_EMAIL
|
| |
|
|
|
|
|
|
|
|
|
|
| |
We have hostname.configure already for setting the hostname, and
nothing in Trove (gitano, LC, etc) should care about the hostname.
They might care about the external name (i.e., TROVE_HOSTNAME),
but that's entirely orthogonal to the system hostname. And some
systems shouldn't have their system hostname set to TROVE_ID. For
example, git.baserock.org's TROVE_ID is baserock, but the system
hostname shouldn't be that (it's currently firehose1).
|
|
|
|
|
| |
Minions load its configuration from /etc/lorry-controller/minion.conf
and the file wasn't been generated at deploy time.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add install-files to the list of configuration extensions, on general
principles. It costs nothing to have it, and is a nice way to install,
say, ssh keys.
In trove.configure, support the optional LORRY_CONTROLLER_MINIONS
configuration key to decide how many Lorry Controller MINION instances
may run at the same time. It has a default in case it is not set.
Create /etc/lorry-controller/webapp.conf. This needs to have some
information from the deployment, so it can't be just installed by
the Lorry Controller chunk.
|
|
|
|
|
|
| |
Since we added releases-repo-migration.sh and a systemd unit to call it,
and both of these need to know the Trove ID, add them to the command that
performs substitutions in trove.configure.
|
|
|
|
|
|
|
| |
This reverts commit 3e73cebdb2ef8df99f416696d040e797b856cd24.
This patch was accidentally pushed without review as part of the
previous branch.
|
|
|
|
|
| |
This is necessary for git.baserock.org, whose TROVE_PREFIX and TROVE_ID is
'baserock' but whose hostname certainly isn't.
|
|
|
|
|
|
|
|
|
| |
This makes it easier to merge changes between /etc when doing upgrades.
Lorry and Mason config goes into local-config.git, not in /etc, so this
still needs to be done at first boot when we can run Gitano commands on
the target. For this reason the /etc/trove-setup.sed file is still
stored in the resulting system.
|
|
|
|
|
|
|
|
| |
The generated /etc/trove-setup.sed script, used to set up site
configuration for a deployed Trove on first-boot by trove-early-setup,
did not specify the 'g' (global) flag which meant that lines with more
than one occurance of a placeholder such as ##TROVE_HOST## were not set
correctly.
|
|
|
|
|
|
| |
It's hard to spot the 'reboot' command at the bottom of the list of
commands and it's a non-standard thing to do. This change should make it
more obvious what goes on during Trove first boot.
|
|
|
|
|
|
| |
If /var is a subvolume then we cannot access it until local-fs.target
has completed, and certainly we can't expect it to be around when systemd
is deciding what units to execute.
|
|
|
|
| |
The trove-early-setup configuration tool is for initial deployment only.
|
|
|