Skip to content

v0.14.0

Choose a tag to compare

@scotwells scotwells released this 10 Nov 23:46
b6e1f7e

Private DNS can run a control plane per region, replicate through a federation hub and share region-written inputs, all opt-in. Serving members drain before they stop, stay healthy through startup, takeovers and record withdrawals, and send far fewer acknowledgements. Private zones no longer need a zone class.

Important

Upgrading from v0.13.0

  • Changes to lease and outbox handling mean old and new control-plane replicas must not share a publication store. Scale the Private DNS control plane to zero, then upgrade, then scale it back up. Don't roll it.
  • Deploy the CRDs from this release before the new control plane starts. dnspublicationmanifests, dnsregistrations, dnsrecordcontributions and dnszones gain fields, and an older CRD would prune them.
  • The DNS Admin role no longer grants write access to dnsnamingpolicies and dnsregistrations; services own them. (#480)
  • Serving members can now drain (agent.healthCheck.drainSeconds). Releases before this one reject that field, so remove it before rolling back.

What's new

  • A Private DNS control plane per region, opt-in with regions. Each region keeps publishing while another is down. (#342, #477)
  • Replication through a federation hub with the new private-replicator role. Control planes can also read inputs from their own cluster (projectDiscovery.local), and regions can exchange the inputs they write over NATS (feed). (#464, #467, #468)
  • Serving members drain before they stop. A member closes its health port and keeps answering for drainSeconds. (#473)
  • Node members can join by pod name pattern (memberPatterns, memberIDFrom). (#499)
  • Broker clients can authenticate with a certificate alone, and a broker password file is reread on every reconnect. (#506, #494, #497)
  • Zone class is optional for private zones. (#479)
  • A managed zone is kept while a network uses it, enforced by the Private DNS admission handler for managed zones. (#491)
  • Networks are limited to 256 zone associations. (#492)

Reliability and performance

  • Acknowledgement renewals no longer trigger full agent reconciles, and long-withdrawn zones stop being renewed. (#512, #514)
  • Member readiness uses per-member leases instead of manifest rewrites, and outbox exports no longer conflict on manifests. (#510, #507)
  • Members wait for their local daemons before probing contexts, avoid reload storms on control-plane takeover, and stay healthy while a record withdrawal installs. (#513, #516, #517)
  • A failing public resolver no longer withdraws DNS contexts. (#518)
  • Unhealthy DNS server pods are removed from their Service. (#521)
  • Project status reads from the watch cache, and the serving checkpoint loads and saves one entry at a time. (#511, #509, #483)
  • Publication acks from node members are settled instead of retried. (#495)

Fixes

  • Cross-zone CNAMEs are followed over UDP. (#490)
  • Private zones are withdrawn from regions they leave. (#477)
  • Project teardown can delete managed zones. (#496)
  • Service-owned DNS resources stay out of DNS Admin. (#480)
  • Remaining Private DNS update-safety and review findings are closed. (#481, #482)
  • How a DNS process exited is logged before it restarts. (#475)

Docs, dependencies and CI

  • The Private DNS API contract is documented. (#476, #487, #493)
  • Go module updates, including a security update to klauspost/compress. (#454, #515)
  • Dashboards show answer and pass times as heatmaps. (#486)
  • CI and E2E cache, dedup and recovery-test work. (#459, #461, #463, #466, #470, #478, #489, #442, #443)

Changelog

  • b963082 Keep public recursion failures from withdrawing DNS contexts
  • 0d35db8 Merge branch 'feat/private-dns-karmada-replication' into feat/private-dns-local-cluster-mode
  • 7bb5b12 Merge branch 'feat/private-dns-karmada-replication' into feat/private-dns-local-cluster-mode
  • b76568b Merge branch 'feat/private-dns-karmada-replication' into feat/private-dns-local-cluster-mode
  • 5e48a2e Merge branch 'feat/private-dns-karmada-replication' into feat/private-dns-local-cluster-mode
  • d01c122 Merge branch 'feat/private-dns-karmada-replication' into feat/private-dns-local-cluster-mode
  • 0a6a811 Merge branch 'feat/private-dns-local-cluster-mode' into feat/private-dns-cross-region-feed
  • 3cbd730 Merge branch 'feat/private-dns-local-cluster-mode' into feat/private-dns-cross-region-feed
  • d756d85 Merge branch 'feat/private-dns-local-cluster-mode' into feat/private-dns-cross-region-feed
  • 690f71b Merge branch 'feat/private-dns-local-cluster-mode' into feat/private-dns-cross-region-feed
  • 470cd38 Merge branch 'feat/private-dns-local-cluster-mode' into feat/private-dns-cross-region-feed
  • d2a97b5 Merge branch 'feat/private-dns-local-cluster-mode' into feat/private-dns-cross-region-feed
  • 6f69228 Merge main into fix/private-dns-region-withdrawal
  • 2787f86 Merge pull request #342 from datum-cloud/feat/regional-control-planes
  • 77d5c14 Merge pull request #442 from datum-cloud/renovate/actions-cache-6.x
  • 848c0f8 Merge pull request #443 from datum-cloud/renovate/major-github-artifact-actions
  • b59deb2 Merge pull request #454 from datum-cloud/renovate/go-modules-minor-patch
  • 243171f Merge pull request #459 from datum-cloud/codex/dns-ci-workflow-efficiency
  • d0616cb Merge pull request #461 from datum-cloud/codex/dns-ci-dedup-followup
  • 1e80b09 Merge pull request #463 from datum-cloud/codex/dns-ci-cache-inputs
  • 96bd12f Merge pull request #464 from datum-cloud/feat/private-dns-karmada-replication
  • 62d2fb3 Merge pull request #466 from datum-cloud/codex/dns-conditional-disk-cleanup
  • 90e9249 Merge pull request #467 from datum-cloud/feat/private-dns-local-cluster-mode
  • caa5285 Merge pull request #468 from datum-cloud/feat/private-dns-cross-region-feed
  • 7efed7c Merge pull request #470 from datum-cloud/codex/dns-shared-build-cache
  • 8c89de4 Merge pull request #473 from datum-cloud/feat/serving-health-port-drain
  • 0bbf1fb Merge pull request #475 from datum-cloud/fix/log-dns-process-exits
  • b692350 Merge pull request #476 from datum-cloud/docs/private-dns-api-contract
  • b6fdbda Merge pull request #477 from datum-cloud/fix/private-dns-region-withdrawal
  • 99ab20a Merge pull request #478 from datum-cloud/test/private-dns-recovery
  • 2795d6e Merge pull request #479 from datum-cloud/fix/private-zone-class-optional
  • 80b55c9 Merge pull request #480 from datum-cloud/fix/dns-admin-private-writes
  • c0eda6b Merge pull request #481 from datum-cloud/fix/private-dns-update-safety
  • f5a868d Merge pull request #482 from datum-cloud/fix/private-dns-review-findings
  • bd2bc3d Merge pull request #483 from datum-cloud/test/serving-checkpoint-memory
  • 4f721b9 Merge pull request #486 from datum-cloud/claude/project-thread-sqyq71
  • cbb2c19 Merge pull request #487 from datum-cloud/docs/private-dns-contract-wording
  • 778eb50 Merge pull request #489 from datum-cloud/codex/dns-config-delivery
  • 2a19c90 Merge pull request #490 from datum-cloud/fix/private-dns-cross-zone-cname-udp
  • 7e0126c Merge pull request #491 from datum-cloud/fix/refuse-managed-zone-delete
  • cc332ec Merge pull request #492 from datum-cloud/fix/private-dns-zone-association-limit
  • c42e49f Merge pull request #493 from datum-cloud/docs/contract-cname-fixed
  • 5d9a9f7 Merge pull request #494 from datum-cloud/fix/reread-broker-password
  • 64ac92c Merge pull request #495 from datum-cloud/fix/drop-node-publication-acks
  • af3d80b Merge pull request #496 from datum-cloud/fix/managed-zone-project-teardown
  • a83e89e Merge pull request #497 from datum-cloud/fix/log-unreadable-broker-password
  • 75e98a4 Merge pull request #499 from datum-cloud/feat/elastic-node-members
  • 484fc12 Merge pull request #506 from datum-cloud/feat/nats-client-cert-reload
  • 3100769 Merge pull request #507 from datum-cloud/fix/outbox-manifest-write
  • 83637a9 Merge pull request #509 from datum-cloud/perf/stream-checkpoint-load
  • dde8086 Merge pull request #510 from datum-cloud/fix/member-readiness-lease
  • fdb57b5 Merge pull request #511 from datum-cloud/feat/event-driven-project-status
  • a2e4134 Merge pull request #512 from datum-cloud/fix/agent-ack-reconcile-churn
  • 8551df8 Merge pull request #513 from datum-cloud/fix/member-startup-verify-gate
  • 7e63811 Merge pull request #514 from datum-cloud/fix/agent-tombstone-expiry
  • ee22087 Merge pull request #515 from datum-cloud/renovate/go-github.com-klauspost-compress-vulnerability
  • 7501a37 Merge pull request #516 from datum-cloud/fix/member-reload-storm
  • cd5a4a0 Merge pull request #517 from datum-cloud/fix/withdrawal-keeps-readiness
  • 37c40a7 Merge pull request #518 from datum-cloud/fix/public-recursion-containment
  • b6e1f7e Merge pull request #521 from datum-cloud/fix/519-probes
  • d44d344 Merge remote-tracking branch 'origin/main' into codex/dns-ci-cache-inputs
  • 3b864ca Merge remote-tracking branch 'origin/main' into feat/elastic-node-members
  • 5bcb8f7 Merge remote-tracking branch 'origin/main' into feat/regional-control-planes
  • 321e534 Merge remote-tracking branch 'origin/main' into fix/log-dns-process-exits
  • e69c8c5 Merge remote-tracking branch 'origin/main' into fix/member-startup-verify-gate
  • 4396b9a Merge remote-tracking branch 'origin/main' into fix/private-dns-update-safety
  • aaf7bac chore(deps): update actions/cache action to v6
  • 97e0e1a chore(deps): update actions/upload-artifact action to v7
  • e260713 chore(deps): update module github.com/klauspost/compress to v1.20.2 [security]
  • be65d63 ci: Cache shared compilation before image assembly
  • 9496770 ci: Cover rule removals and superseded E2E runs
  • 2d75c46 ci: Deduplicate generation and harness checks
  • e89f9d1 ci: Integrate current main for validation
  • 9ed6d6e ci: Integrate merged workflow optimizations
  • 65976a6 ci: Keep operator publishing independent
  • 900068a ci: Pass only image digests between build steps
  • e20fbbf ci: Reject missing test results at the gate
  • 3612188 ci: Retain validation on release tags
  • 7d0735e ci: Reuse image inputs and overlap cache export
  • 0d6ab62 ci: Reuse validation and image build work
  • 48bb2d6 ci: Skip disk cleanup when headroom is sufficient
  • b6f23fd ci: deliver fixture DNS configuration sooner
  • f24f25e ci: share the node loopback address
  • 08cf80e feat: Drain serving members before they stop
  • e403942 feat: Exchange region-written DNS inputs between regions over NATS
  • 3e16b3a feat: Keep a region-written object's UID on every copy
  • a173982 feat: Label mapped namespaces for the networking projection's placement
  • 9da4818 feat: Let a broker client authenticate by cert alone
  • cd48e51 feat: Let node members join by pod name pattern
  • 9c22453 feat: Read private DNS inputs from the control plane's own cluster
  • aa49d13 feat: Record each hub copy's origin UID
  • 4a38dd0 feat: Replicate private DNS objects through a federation hub
  • f8c6be2 feat: Run a private DNS control plane per region
  • 735d309 feat: make zone class optional for private zones
  • ec71fe3 feat: show answer and pass times as heatmaps
  • 7c48b66 fix(deps): update go modules (minor and patch)
  • ab3a176 fix(internal-dns): follow cross-zone CNAMEs over keyed UDP
  • 2f4b70c fix(private-dns): refuse zone associations past 256 per network
  • 60a062b fix(private-dns): write a zone association's condition once per pass
  • 79e8eb4 fix: Accept a peer's inputs only in projects it serves
  • 7011a87 fix: Add readiness probes to the DNS server sidecars
  • e32d516 fix: Build the private replicator with the operator image
  • 1985d0a fix: Close readiness gaps around withdrawals and gating
  • 93a5921 fix: Close remaining private DNS review findings
  • c5fe8c2 fix: Close review gaps in local-cluster mode
  • ddb4785 fix: Close review gaps in private DNS replication
  • 3bcca9a fix: Close review gaps in regional control planes
  • 2b56522 fix: Close review gaps in the cross-region feed
  • 3b4da39 fix: Close the remaining Private DNS update safety gaps
  • 17dda51 fix: Cover untracked and removed shards in withdrawal
  • a223555 fix: Drop unused labels from member leases
  • 7ca84cd fix: Fail over public upstreams on the failures that matter
  • b484b5d fix: Give parsed zones an upstream for CNAMEs
  • 7868888 fix: Guard managed zone release and withdrawal
  • f7cb446 fix: Harden health port and drain from review
  • def8a9f fix: Hold zones only in regions that serve them
  • 16d239a fix: Ignore exit log write errors explicitly
  • c99a482 fix: Install a project's cache only once its zones and registrations sync
  • d13f84b fix: Isolate member lease read failures
  • af3bb6c fix: Keep a managed zone while its network uses it
  • bdec1a0 fix: Keep approval beside withdrawn and pending contexts
  • ce001d8 fix: Keep console clients in the update's process group
  • f114e42 fix: Keep failed daemon updates logged as errors
  • 3f15cf9 fix: Keep gate stops and crash loops out of the exit log
  • 5ae997d fix: Keep managed zone owners and label fixed
  • 392a994 fix: Keep memory consistent when retirement fails
  • c943521 fix: Keep pattern members required until retired
  • 2a99f8f fix: Keep reconnecting through rotation auth errors, reload client certs
  • 489d152 fix: Let project teardown delete managed zones
  • ef69d99 fix: Log a broker password file that stops being readable
  • afa6e7a fix: Log how a DNS process exited before restarting it
  • c946280 fix: Log why a broker connection dropped, return a copy of the cert
  • f103ddd fix: Name upstream addresses without shadowing the import
  • 0ad13df fix: Normalize cache policy and update lane tests
  • 8d76d61 fix: Probe once after an expired local wait
  • 3fd8d0d fix: Read hub copies with their origin's generation
  • 43c1ce4 fix: Read the broker password file again on each reconnect
  • 5dd62a0 fix: Refuse a peer's claim on another object's identity
  • 0ac5779 fix: Remove a peer's copies when its access goes, and keep KV purges
  • 211d9a0 fix: Renew member readiness leases, not manifests
  • f42178e fix: Report written upstream answers as handled
  • 9cafe6f fix: Resolve references between hub copies by their origin's UIDs
  • 7687086 fix: Settle an outbox only after marking the export
  • 5f90090 fix: Settle publication acks from node members instead of retrying
  • 610e4b6 fix: Size publication requirements as encoded
  • 0998251 fix: Stay ready while a withdrawal installs
  • f80d539 fix: Stop node members from sending rejected publication acks
  • bbaa8b7 fix: Stop outbox exports conflicting on manifests
  • d34f0c6 fix: Stop renewing acks for long-withdrawn zones
  • 8b140c0 fix: Take the withdrawal lock only when installing
  • f032177 fix: Track pattern members that ack stale revisions
  • 4e76178 fix: Wait for local daemons before context probes
  • b3f27b0 fix: Withdraw private zones from regions they leave
  • 192079d fix: exclude CI caches from the build context
  • 489ce8f fix: keep heatmap colours and layout honest
  • 5db761d fix: keep service-owned DNS resources out of DNS Admin
  • 2996ca5 fix: tolerate zones without a class in metrics
  • 7776515 perf: Load the serving checkpoint one entry at a time
  • 24f52bc perf: Read project status inputs from the watch cache
  • d392be3 perf: Save the serving checkpoint without a second copy of it
  • d293071 perf: Skip reconcile on same-fence ack refreshes
  • ecf8a1a refactor: Resolve the admission project client once
  • ef10244 style: Group project discovery imports
  • 11fcf5e style: Wrap long replicator setup lines
  • 76cb00f test(private-dns): cover zone limit ordering by acceptance time
  • 2104ab9 test(private-dns): preallocate zone limit test fixtures