This modified script replaces adds a wait condition on startup before attempting
to interact with the database. This is done with 'mariadb-admin ping' run in a
loop at 1s intervals, ultimately timing out after 30s if unsuccessful. Other
minor updates to the script print additional info the docker logs during
startup.
v2.0 was built using a demo database that should not have been present. v2.1
uses the same application, but should start with a clean database. This change
bumps the version to 2.1 and resets the database if the currently installed
version is 1.* or 2.0.
The script `install.sh` largely copies the template from the hosted-app-utils
project, with some key changes:
1. replace template parameters (`{{VAR}}`) with hard-coded values
2. define variables for org, app, install root, etc
3. if old version detected, remove data folder before installing update
The modified script will:
1. delete v1 data
2. update normally when v2 or newer installations are present
The wkhtmltopdf library does not work with the latest version of alpine. To
workaround this, a known working image based on an older alpine image is pulled
in and relevant libraries copied to the target image.
The `snappy` package required a few updates to support this, including setting
the wkhtmltopdf binary path explicitly and to enable access to local files
(disabled by default).
The table `gender_based_violences` was missing a `triage_id` field, causing an
exception to occur in some cases when selecting the triage button. This change
alters the schema of the table to ensure the column is always present.
The latest images incorporate all changes previously added by the dockerfiles.
The arm image however is still missing/unreliable from the registry and must be
built independently.
Two bugs are fixed by this commit. First, a typo in the format string used write
resource usage data to the systemlog caused the systemlog to (correctly) reject
the string as invalid json. The second fix is to handle a case where a .env file
is present, but does not define the APP_CONTAINER variable. In this case the
container name is left empty. This is not an immediate issue, but would become
problematic in the future if additional containers were to be added.
While 7 is the default number of characters used to abbreviate a commit hash,
this is not (by default) constant and different instances of the same repo may
generate different version strings as a result. This commit sets this length
explicitly to 7 in both package.sh and build-image scripts.
An updated set of source files and initialization was provided by streamline to
address issues observed during initial testing. These files have been updated in
order to generate a new set of app images.
The notion of application "instances" is dropped in favor of a simpler,
single-instance approach. Default values are provided for all configuration
variables so allow the app to run without additional manual configuration.
A health check and auto-heal script are added to automatically restart the
application in case it becomes unavailable during operation.
- compress images with xz by default (and decompress on install)
- add help menu and command line argument parsing to package.sh
- display packaging status if `pv` command available
The official image from streamline does not currently work for the arm64
platform. As a temporary measure, the source code and docker build scripts
have been lifted from the official images and are used to build locally.
Some additional modifications are made to reduce overall image size, these
are documented in docker/README.md
Dockerfiles are updated with two key changes:
- use streamline's official images as a base
- replace two files related to mysql configuration
A README.md file is added to the docker folder which details the changes, but
the key difference is to limit binary logs to only the `streamline` database.
This prevents additional system databases from being included, which the
database sync tools do not (yet) support.
An unrelated bugfix corrects the docker-compose image tag, specifying 'latest'
instead of a previously hard-coded version.
interrupts were not properly handled by the entry script, and could leave
containers active. This is fixed by running stop in a cleanup function, called
on exit.
Eventually, images should not be transferred as part of the application package,
using a more efficient mechanism instead. Other cases may already exist where
packaging without the image is preferrable.
When the `PACKAGE_IMAGES` variable is not set to 1 (default: 1), images will not
be included in the package. The installation script now handles the case where
the `images` folder is missing altogether.
An additional config file `instances.conf` is added to define instance names and
port numbers. The existing `app.conf` was also modified to include an
`instances` field and to move node id and location under this new object. A
`.env` file is used to manage configuration for each instance.
Additional scripts are added: `start`, `stop`, `monitor` - see docs for
description. The main entry script is also modified such that it will run until
any service stops, then brings down the remaining services and exits.
the package format was slightly off - there should have been a folder in the
root of the archive wrapping the package contents. This is now implemented.
Further configuration options are made available to support running multiple
instances of the app on the same machine, including the use of docker sibling
containers. `README.md` and `template.env` are updated accordingly.
The database port is no longer exposed as this is not required to access and
manage the app at this time.
- A bug in `entry` caused the script to not change into the directory correctly.
This was corrected by marking the script as bash rather than sh.
- `install.sh` now stops a running service before updating
- `docker-compose.yml` now assigns a consistent container name for the app
- `docker-compose.yml` now uses environment variables or a `.env` file to
assign container name and port mappings if available
- updated readme file with better instructions
The packaging script will create a new tar.bz2 archive containing an install
script, a set of docker images, and a root folder with files to be installed.
The root folder includes a systemd service, a docker compose file, an entry
script, and a template config file.
On installation docker images are loaded onto the target system, install files
are copied, then systemd services are installed.
alec.turner