The release for Project Clearwater sprint “Fingolfin” has been cut. The code for this release is tagged as release-123 in GitHub.
This release we’re upgrading etcd to the latest and greatest version.
This version isn’t directly compatible with the etcd used in the previous release, so after you’ve upgraded all your nodes you’ll need to recreate your etcd cluster following the instructions at http://clearwater.readthedocs.io/en/latest/Handling_Multiple_Failed_Nodes.html. Until you do this the etcd processes on your nodes will report failures, and you won’t be able to easily upload changes to shared config or add/remove nodes.
The release also includes the following bug fixes:
- Some calls fail if java fails on 1 of 3 vellum nodes.
- There is no equivalent to mark_node_failed that can be used to remove a Vellum node in the other site from the Cassandra cluster
- Queue Manager doesn’t log on being stopped
- Incorrect case for alarm severity in Sprout communication error alarms.
- Spurious “Error impistore.cpp:664: Failed to write IMPI for private_id email@example.com” log on CAS failure
- upload_json (and therefore all our JSON uploaders) make it easy to upload an invalid file
- We should write better/more noticable logs on removing the data directory on etcd startup
- Clearwater diags monitor should pull off the healthy_etcd_members file
To upgrade to this release, follow the instructions at http://docs.projectclearwater.org/en/stable/Upgrading_a_Clearwater_deployment.html. If you are deploying an all-in-one node, the standard image (http://vm-images.cw-ngv.com/cw-aio.ova) has been updated for this release.