- CRITICAL : change rest.properties config
- contribute back to geonode-project
- CRITICAL : see if Geoserver authkey tokens expire (even when the key is deleted from the database, it's still possible to use it until manually clicking "sync user/group service". It looks like it's some cache, but I don't know if it expires. Maybe we need to use webservice instead of user property...)
- fix updatelayerip on startup (currently creates a mess in links when host/port changes and deletes custom thumbnails)
- make monitoring module work (currently it's disabled because of some exception during startup)
- check if everything is ok with auth_keys (it seems Geonode uses expired keys...)
- tweak nginx settings (gzip output, cache, etc...)
- use alpine for django as well
- migrate to spc repositories instead of olivierdalang
- see if we can have geoserver exit on error, in not at least implement proper healtcheck
- keep a version marker in the geodatadir directory in case of updates to the datadir
- set more reasonable logging for geoserver
- add at least some basic integration test to travis
- see if we can setup something for backups on local filesystem
WARNING YOU CANNOT UPGRADE FROM 0.0.x to 0.1.x
YOU NEED TO DO A FRESH INSTALL AND MANUALLY TRANSFER THE DATA
- improved nginx config (gzip and expiration header)
- targetting future 2.10
- removed elastic search container (it was unused anyways)
- removed postgres login hack and using instead Geonode-Geoserver OAuth mecanism
- prebuilt geodatadir used again and master password procedure simplified
- added django healthcheck
- if https is enabled, force redirection to https host (as geonode doesn't support multiple domain names/relative installs)
- django secret generated automatically
- undo admin users disabled again
- revert using 2.6.x branch (because of side effect - login taking ages)
- use Geonode's Geoserver .war build instead of starting from vanilla
- fix thumbnail generation (uses a custom release of Geonode)
- django admin users are again disabled on restart (so we can keep only 1 superuser)
- added travis integration test (try to deploy django then tries to create an user, upload a layer, get the thumbnail and get a tile of the layer)
- changed rclone configuration (you must now provide rclone conf file)
- removed syncthings
- make http(s) ports parametrable in case a port is already busy
- various fixes (broken pip dependencies, wrong fix for geoserver proxy, ssl certificate refreshing)
- siteurl set using HTTPS_HOST or HTTP_HOST (instead of "/" which isn't supported)
- use custom build of geonode (with some fixes not upstreamed yet)
- geoserver master password reset is cleaner (programmatically reset the password from initial datadir before first launch)
- support empty HTTP_HOST or HTTPS_HOST
- geosever 2.12.1 => 2.12.2
- cleaned up env vars
- upgrade should work
- improve nginx<->letsencrypt (nginx can work without letsencrypt service)
- put django in main directory (so it's more clear for deploy builds)
- removed rancher template from repo
- removed entryponts and command from django image to prevent what looks like a bug in rancher where empty entrypoint in docker-compose isn't taken into account
- added a second backup service using RClone (the idea is to test both syncthings and rclone then choose one)
- we don't rely on an initial geodatadir anymore, instead we start from scratch, launch geoserver once, then do our modifications
- added a backup service using Syncthings
- fix bug with rancher resolver on rancher
- allow to disable/test let's encrypt using env variables
- we use geonode users/groups table directly for geoserver's authentication
- have ssl working online
- use env variables / secrets where applicable
- publish on git and autobuild images
- make docker deploy work again