- 5月 28, 2020
- 5月 27, 2020
-
-
由 Ismael Juma 创作于
We did a poll to check if people would prefer to keep the minor version number under 1000 by bumping the major version and the result was yes (7-3). Reviewers: Dhruvil Shah <dhruvil@confluent.io>
-
由 Manikumar Reddy 创作于
-
由 David Jacot 创作于
-
由 Dhruvil Shah 创作于
-
由 Steve Rodrigues 创作于
* Define scheduleAddBrokers API for the DataManager class. * Test broker add detection in the controller. * Detect new brokers in the KafkaController::onBrokerStartup callback, and call into SBK with the list of new broker IDs.
-
由 Sanjana Kaundinya 创作于
-
由 Rajini Sivaram 创作于
CONFLUENT: Add PowerMockIgnore for Scram API in RBAC test to fix classloader issues in subsequent tests (#1790) SCRAM providers loaded using PowerMock classloaders can cause failures in subsequent tests that don't use PowerMockRunner since a mixture of classes loaded by different classloaders will be used and this breaks callback handlers. This commit adds PowerMockIgnore for SCRAM API classes. Reviewers: Sanjana Kaundinya
-
由 Milo Simpson 创作于
- added new ClusterRegistry ResourceType that ClusterAdmin has access to - SystemAdmin role gets access to ClusterAdmin via it's All/All grant - added special casing logic to ConfluentProvider.isSuperUser method to allow MDS broker super.user to be able to CRUD all ClusterRegistry data Also - took a stab at documenting the Roles and Resources types in the json file that defines them
- 5月 26, 2020
-
-
由 Brian Byrne 创作于
- 5月 25, 2020
-
-
由 Jeff Huang 创作于
Reviewers: Manikumar Reddy <manikumar@confluent.io>, Ismael Juma <ismael@juma.me.uk>
-
由 Stanislav Kozlovski 创作于
This patch revises the access strategy used for the Executor. The remove broker feature will need to reserve the Executor prior to making use of the class (during broker shutdown and plan computation) in order to ensure that self-healing (or anything else) doesn't trigger another proposal execution under its feet. Further, because we have the notion of a "critical operation" that overrides others (as per the UX doc), it's useful to have a straightforward way of cancelling ongoing operations/locks and taking over.
- 5月 23, 2020
-
-
由 Lucas Bradstreet 创作于
I noticed that the objectId in the messages logged by ArchiveTask are not the base64 encoded variants. I think we might as well print the whole metadata using its toString method to avoid this problem. Example of this problem: ``` Completed UploadInitiate(objectId: bb38ea89-22ca-499b-a4a6-c40daf74df8c, baseOffset: 0, endOffset: 1919380]) for mntPrHSbTgGCQ-2WfTbz2A-lkc-pgo8m_di_kafka_presence_interval_processor_V1_prod-backfill-metroid-HISTORY_TABLE-changelog-6 in 18ms ```
-
由 Lucas Bradstreet 创作于
Log out additional fetch information when the TierFetcher encounters an exception.
-
由 Rigel Bezerra de Melo 创作于
-
由 Lucas Bradstreet 创作于
Adds support for tiered storage to the downgrade system test. This will be especially helpful for tiered storage format changes. Unlike the normal downgrade test, we defer consumer fetches until after the downgrade has completed, as we wish to hit the tiered section of the log. I have also added CP 5.4 and CP 5.5 into the non-tiered storage test matrix.
-
由 Brian Byrne 创作于
-
由 Kowshik Prakasam 创作于
-
由 Kowshik Prakasam 创作于
-
由 Brian Byrne 创作于
- 5月 22, 2020