Release/2019 05 02

[ChrisPenner]

Documentation changes

• Improved Cassandra documentation in docs/README.md
• Improved documentation on SCIM storage in docs/README.md
• Improved documentation on SCIM Tokens in docs/reference/provisioning/scim-token.md

Bug fixes

• Sanitize metric names to be valid prometheus names in metrics-core
• Add missing a .git suffix on gitlab dependencies in stack.yaml
• Time bounds checks now allow 60s of tolerance; this is helpful in cases of drifting clocks ( Add hard-wired 60 sec tolerance to all time bounds checks. #730)

Features

• Services now provide Prometheus metrics on /i/metrics
• Garbage Collection and memory statistics are available alongside other prometheus metrics

Internal Changes

• Alpine Builder is no longer built with --profile
• SCIM users now have an additional wire-specific schema attached.

Changes (potentially) requiring action

• /i/monitoring is DEPRECATED. Please use prometheus metrics provided by /i/metrics instead.
• On password reset the new password must be different than the old one
• Stern is now available as a new tool for performing administrative tasks via API (Stern as an extra tool #720)
• SCIM handler errors are now reported according to SCIM error schema (In SCIM handlers, format (almost) all errors using SCIM error schema #575)

[fisx]

• I can't find In SCIM handlers, format (almost) all errors using SCIM error schema #575, Add hard-wired 60 sec tolerance to all time bounds checks. #730 in the release notes? At least Add hard-wired 60 sec tolerance to all time bounds checks. #730 is quite relevant.
• I liked the fact that the release notes referenced the relevant PRs in the last releases. We can skip it this time, but it would to have that back in the next release.

[ChrisPenner]

@fisx 575 was listed under internal changes; I moved it to the public changes :)

Sorry I missed 730! It's been added 😄

[fisx]

@jschaul any reason we shouldn't include #725?

[jschaul]

@jschaul any reason we shouldn't include #725?

Don't release untested changes. That PR isn't merged because the upstream way of building docker images changed and needs updating/testing first (which I only saw after I made that PR).