Repository navigation
Conversation
Signed-off-by: Baylem <zelathis@gmail.com>
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the
|
…rld data Signed-off-by: Baylem <zelathis@gmail.com>
Signed-off-by: Baylem <zelathis@gmail.com>
CI Report - 30db243run 43162327533 - attempt 1 0 failed - 29 passed - 4 skipped - 127m49s elapsed Test selection: partial - per-suite table Coverage
Changed
Slowest jobs
e2e buckets run
|
Signed-off-by: Baylem <zelathis@gmail.com>
…larations Signed-off-by: Baylem <zelathis@gmail.com>
…indow Signed-off-by: Baylem <zelathis@gmail.com>
…l backups Signed-off-by: Baylem <zelathis@gmail.com>
Signed-off-by: Baylem <zelathis@gmail.com>
|
Hey, could you please add changelog entry? |
Signed-off-by: Baylem <zelathis@gmail.com>
|
Added the requested unreleased changelog entry. CI passed on a0891c6b. The diff through 955171aa changes only the changelog entry and its index link. I canceled the repeat CI run because this update changes only Markdown. |
…obe parents Signed-off-by: Baylem <zelathis@gmail.com>
Signed-off-by: Baylem <zelathis@gmail.com>
…cation Signed-off-by: Baylem <zelathis@gmail.com>
Closes #870. Parent tracker #867. ARK: Survival Ascended and Project Zomboid now declare stopped-copy backup guarantees in the upstream module catalog. Module PR #59 · Merge 2a7bb6c8
Gameplane’s shared catalog delivery prerequisite is #926. The maintainer updated its gitlink to
f9e55aa1after the ARK: Survival Evolved 2.0.1 safeguard merged. This pin includes both this backup correction and the Palworld telemetry change. Installations using OCI catalogs need a release containing this revision. Catalog pin and CI, #926The operator records pod-bound verified freeze, acknowledged flush with writes still enabled, and drained stopped capture separately. Historical unqualified hook successes remain restorable and report unknown consistency. The Minecraft Java checkpoint adapter, validation and agent protocol express flush-only without creating a resume obligation. CI exercises restored Minecraft world and scoreboard markers, a client join, another save, and SQLite integrity after each of two stopped Restic copies.
The durability (#899) and probe (#889) changes have merged. This review now stacks on #900 via
stack/870-backup-and-probesfor its existing one-request webhook fixture fix. The arm64 operator run against master exposed the BusyBox sink closing a connection that the real notification client reused. In a disposable reproduction, 150 of 300 requests failed on the original fixture. With theConnection: closechange from #900, all 300 requests arrived exactly once. Commit8cff7b87retains the same 50 backup files with 473 additions and 61 deletions. Merge #900 before retargeting #907 to master. Failed operator job · Existing fixture fix · Webhook rerun. Real production recovery evidence on other engines depends on separately qualified engine adapters.On
8cff7b87, the arm64 operator test passed, as did the backup and offline-recovery drills. The amd64 placement drill independently failed before applying the quota because its Helm release and its disposable tenant shared a namespace. #886 now renders this fixture using the control-plane release namespace and keeps both the admission binding and the quota scoped to the disposable tenant. The overlapping namespace still fails admission validation in Helm. #907 consumes this change at30db243f; its scoped diff remains 50 files, 473 additions and 61 deletions. Passing arm64 webhook test · Placement failure · Shared fix · Full backup workflow on30db243f.Recovery and placement on signed head
30db243fd92687365561e88395962fccb27d2a2bThe workflow finished with 73 successful jobs and four conditional skips. The arm64 operator suite also passed on this head. Final CI report · Arm64 operator
The backup preparation and same-node copy drill passed in 113.69 seconds. Two independent stopped Restic snapshots restored a SQLite database with
PRAGMA integrity_checkreturningokand the player score still 17 (190.94 seconds). The multi-volume drill passed its suspended, idle, missing-claim and changed-path cases in 506.11 seconds. Same-node copy · SQLite and multi-volume restoresMinecraft Java 1.21.4 accepted the client after Restic restored the captured world marker and player score 17. The restored game accepted score 18 and flushed it to disk. The live drill took 385.73 seconds. Minecraft restore, rejoin and save
All three placement drills passed. Two workloads recovered after drain in 3 and 4 seconds with the same world markers and volume identities. The leaking tenant reached its 32 MiB limit with the Pod cgroup OOM counter increasing from 0 to 1 and the kill counter from 0 to 2. The neighboring Minecraft server recorded p95/p99 tick durations of 0.3/2.8 ms under load and 0.1/0.2 ms after the kill. The Restic backup and repository read on the surviving worker took 19.63 seconds. Worker placement and disruption
Recovery qualification on
a0891c6b9fa749a900ca1d7d925e576a6046e022Minecraft Java 1.21.4 accepted the real client before capture and after Restic restored the save. The world marker and player score changed from their captured values during the acknowledged freeze, and changed again after the copy. The restored server returned the original world marker and score 17, accepted score 18 and acknowledged another flushed save. A post-resume marker reached disk through native autosave. The live test completed in 532.71 seconds. Minecraft capture, restore and rejoin
Two independent stopped Restic snapshots each restored a SQLite database with
PRAGMA integrity_checkreturningokand the expected player score of 17 (190.68 seconds). The multi-volume recovery test passed its suspended, idle, missing-claim and changed-path cases (556.46 seconds). ARM64 operator recovery tests