v1.27.0

v1.27.0

Changes

Operator

• Update fdb go bindings to allow copile on MacOS #1858
• Add e2e test to make sure processes in maintenance zone are ignored #1848
• Update docs about resource limits and update defaults to 4GB for a minimal cluster #1847
• Improve the handling of test processes #1846
• Make sure that the PVC and service gets created if a process group gets quickly marked as removal #1845
• Add more logging for upgrade test failure #1843
• Fix flaky migration test setup #1842
• Remove the list call in the remove process groups reconciler and use the get method instead. #1838
• Add failure condition to logging in automatic replacements #1834
• Make the verification of the coordinators more flexible and use the hard limits to validate the locality distribution #1833
• Bump golang.org/x/net from 0.10.0 to 0.17.0 #1832
• Allow e2e test suite to specify downgrade versions #1831
• Increase knob rollout time #1830
• Update docs on fault domain and coordinators #1826
• Add DataLoader to make sure cluster has minimal data #1762
• Initial support for three data hall replication #1651
• Initial doc for replacement buckets #1368

v1.26.0

v1.26.0

Changes

Operator

• Correct the way an empty result is interpreted #1828
• Add coordinator fault tolerance check #1827
• Fix the exclusion logic for Pods that are recreated before being fully excluded #1825
• Let operator replace processes that are excluded but not yet marked as removed #1824
• Improve test stability #1823
• Handle large exclusions better #1822
• Add e2e test for changing a cluster to use DNS in cluster files #1821
• Add first upgrade test with DNS usage in cluster file and correct statement about DNS #1820
• Service missing namespace #1818
• Investigate HA test failures #1817
• Remove the MissingPVC condition for stateless processes #1815
• Add first set of tests for locality based exclusions and fix bug in locality based exclusions #1811
• Make use of errgroup and simplify some testing code #1810
• Refactor code to make use of the process group directly without passing the class and id number down #1807

v1.25.0

v1.25.0

Changes

Operator

• Split upgrade tests into tests with chaos mesh and without #1812
• Split upgrade tests as we hit the timeout for our CI pipeline #1808
• Correct the namespace creation for e2e tests #1806
• Allow to use max timeout for get status and allow to specify the max timeout #1805
• Improve operator lock handling by releasing locks once cluster is reconciled #1803
• Correct the value used for testing #1802
• Make use of the new fault tolerance methods #1797
• Fix the lock ID to return the correct value #1796
• Only set the MissingProcesses condition if the machine-readable status contains at least one process #1794
• Increase timeout for default single cluster upgrade #1792
• Initial design doc for suspending Process Groups #1785
• Remove unused hot-shard tool #1784

v1.24.0

v1.24.0

Changes

Operator

• exposing context #1790
• Fix Issue #1739 #1789
• Update the default e2e FDB versions #1783
• Fix CRD validation for Process Group ID and allow wildcard #1779
• Allow to update the annotations and labels for the FoundationDBCluster resource in tests #1778

v1.23.0

v1.23.0

Changes

Operator

• update goreleaser version, and remove invalid config #1780
• Add namespace parameter for PVC query and move process count validation into common method #1774
• Add e2e test to make sure operator can handle Pods that are stuck in pending #1773
• Remove the client service in the sample #1772
• Document the maintenance mode feature #1771
• Allow to reduce debugging information in operator test framework #1770
• Update goreleaser #1767
• updating ipv6 support in sidecar and service #1754
• Allow to create controller runtime from different context #1745
• Update controller gen #1738

v1.22.0

v1.22.0

Changes

Operator

• Add messages from the client and cluster part in the machine-readable status #1761
• Remove the additional get status call for getting the coordinator set #1759
• Check if the database is available before doing any exclusion checks #1758
• Update fault domain detection for multiple processes in a single Pod #1753
• Verify the processes that are excluded #1752
• Fix get removal mode #1751
• Add label watch perdicate #1744
• Upgrade to go version 1.20 #1740
• Split deployment.yaml config sample RBAC roles #1734

v1.21.0

v1.21.0

Changes

Operator

• Remove logServersPerDisk from v1beta1 API #1742
• Fix linting for status fields that are lists #1737
• Explanation for the number of recoveries that happen during an upgrade #1736
• Disable node access per default in the operator #1735
• Document used FDB ports #1733
• Correct the way how the client knobs are passed down and used #1731
• Replacement logic should ignore process groups that are in maintenance mode #1711
• supporting multiple Log processes per disk #1686

v1.20.1

v1.20.1

Changes

Operator

• Fix exclusion logic multiple processes #1721

v1.20.0

v1.20.0

Changes

Operator

• Correct the way how the e2e test framework checks the username #1717
• Expose methods to create a cluster based on the generated spec #1716
• Add a method to the e2e factory to collect logs for a Pod #1715
• Document upgrades for 7.1+ and add the default log group to the ignore list #1713
• Add some more logging if exclusions check is skipped #1709
• Add data lag metrics from machine-readable status #1708
• Run kill command once without back off to reduce risk of multiple recoveries #1706
• Improve the logging for the cluster controller to make sure all fields are set #1705
• Make sure we wait some time before checking the incompatible connections #1704
• Update chaos-mesh and the controller runtime version #1701
• Make sure mock k8s client implements controller runtime client #1700
• Allow to mock an error in the mock client #1699
• Fix panic in analyze command if cluster is not found #1697
• Fix test case for version compatible upgrades #1696
• Add prefix factory option #1695
• Disable availability check for unstable network test #1692
• Move status checks dedicated package #1691
• Do fault tolerance checks on status object #1689
• Update taint feature design doc and user manual #1632

v1.19.0

v1.19.0

Changes

Operator

• Extend HA upgrade test to check that process groups are only replaced once #1682
• Reduce the size of the process map by selecting only processes in the right DC #1681
• Add logging when the operator detects a new running version #1678
• Update base image for operator #1677
• Run HA upgrade tests on PRs #1676
• Make sure we are not allowing to remove process groups if they have no addresses assigned until the exclusion is skipped #1673
• Make sure we update the sidecar images too for patch upgrades #1672
• Update go to v1.19.10 #1671
• Add better logging for spec hash changes #1670
• Fixes issue #1640 #1669
• Simplify the parsing of the ID number for the process group ID #1667
• BugFix: init faultDomainsWithUnavailablePods before checking ProcessGroups updates #1666
• Update customization.md #1664
• Enable "maintenance mode is on" test in "operator_test.go" #1663
• Initial support for cache status per reconciliation loop #1661
• Take maintenance mode into account while choosing coordinators and read maintenance mode information from special key space #1652
• Fix kind e2e tests. #1630
• Add MaxZonesWithUnavailablePods for the update process. Fixes #1444. #1587