v0.9.0
Known issues
v0.9.0 accidentally assumes that there are no deleted service instances in etcd; which means they exist as service/ID
but lack the service/ID/state
data object; and this breaks adding nodes to existing clusters and breaks the ability to create new clusters. Fix coming in v0.9.1. #10 Oops, sorry!
Breaking Changes
Because sanity-test tests the 'move cluster' feature (see below) it requires a minimum of 4 cells to complete successfully. Installations with less will have to increase the cell instance count to complete the sanity-test.
Improvements
Administrators can create a cluster on behalf of a user into specific cells/vms:
cf create-service dingo-postgresql cluster their-db -c '{"cells": ["10.244.22.2", "10.244.21.7"]}'
Administrators can now "move" a cluster to different cells/vms
cf update-service their-db -c '{"cells": ["10.244.22.2", "10.244.21.7"]}'
The reason the cells
parameter above is "for admins" is the requirement to know about the available/desirable "cells", which is an invisible concept to end users.
Use /admin/cells
to discover available cells:
curl ${BROKER_URI}/admin/cells
Fixes
- Allocation of nodes now correctly assigned into different AZs
- Nodes are prioritized into cells that have fewer nodes
To upload BOSH releases:
curl -s "https://api.github.com/repos/dingotiles/dingo-postgresql-release/releases/tags/v0.9.0" | jq -r ".assets[].browser_download_url" | grep tgz | xargs -L1 bosh upload release --skip-if-exists
Or get URLs for BOSH releases:
curl -s "https://api.github.com/repos/dingotiles/dingo-postgresql-release/releases/tags/v0.9.0" | jq -r ".assets[].browser_download_url" | grep tgz