From 34a7d9703a9ea7050e0e35ae6b1c8835d25f7b14 Mon Sep 17 00:00:00 2001 From: Bryan Burkholder Date: Wed, 7 Feb 2024 11:15:30 -0700 Subject: [PATCH] PR feedback --- .../java/com/slack/kaldb/server/RecoveryTaskCreator.java | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/kaldb/src/main/java/com/slack/kaldb/server/RecoveryTaskCreator.java b/kaldb/src/main/java/com/slack/kaldb/server/RecoveryTaskCreator.java index 17b5b63f9a..dca25b4267 100644 --- a/kaldb/src/main/java/com/slack/kaldb/server/RecoveryTaskCreator.java +++ b/kaldb/src/main/java/com/slack/kaldb/server/RecoveryTaskCreator.java @@ -218,8 +218,9 @@ public long determineStartingOffset( } else if (indexerConfig.getCreateRecoveryTasksOnStart() && indexerConfig.getReadFromLocationOnStart() == KaldbConfigs.KafkaOffsetLocation.LATEST) { - // todo - this appears to be able to create recovery tasks that have a start and end - // position of 0, which is invalid + // Todo - this appears to be able to create recovery tasks that have a start and end + // position of 0, which is invalid. This seems to occur when new clusters are initialized, and is + // especially problematic when indexers are created but never get assigned (ie, deploy 5, only assign 3). LOG.info( "CreateRecoveryTasksOnStart is set and ReadLocationOnStart is set to current. Reading from current and" + " spinning up recovery tasks");