Info

According to some of our tests, the shard draining is not working correctly in 8.2 and 8.3 both. What is failing ? We have a test which performs a transition from shard to dedicated config server. When our agent polls the status, it comes of as completed in 8.0 but not in 8.2 or 8.3. Why is it failing ? (preliminary analysis, server help needed) MongoDB 8.2 & 8.3 Step 1: Initial chunk move succeeds (rotated log, before the loop) 20:57:14  config.rangeDeletions collection created on myClusterShard1 (via oplog) 20:57:23  Starting chunk migration donation — fromShard:"myClusterShard1" → toShard:"config" 20:57:23  Migration succeeded (128ms) ✓ The cluster setup moved a chunk from shard1 → config. That succeeded. But it left orphan documents on shard1 (the donor) scheduled for deletion in config.rangeDeletions. --- Step 2: transitionToDedicatedConfigServer tries to move it back (main log) Now the same chunk needs to go config → shard1. Every single attempt hits: 20:57:34  Starting receiving end of chunk migration  range: {a:5000} → {a:MaxKey} 20:57:34  Waiting for deletion of orphans orphanRange: {a:5000} → {a:MaxKey} 20:57:44  Error during migration: ExceededTimeLimit "Migration failed because the orphans cleanup routine didn't clear yet a portion of the range being migrated that was previously owned by the recipient shard." This repeats every ~11 seconds, 189 times, from 20:57:34 to 21:32:17. --- Step 3: The explicit smoking gun (last lines of log) At 21:32:18 — the very last migration attempt adds a new, even clearer message: 21:32:18  Migration paused because the requested range overlaps with a range already scheduled for deletion range: [{a:5000}, {a:MaxKey}) And the final changelog entry: 21:32:17  moveChunk.to  note:"aborted"

Top User Comments

Steps to Reproduce

Initiate the transitionToDedicatedConfigServer step.