Loading...

XML

Word

Printable

JSON

Assigned Teams:

Catalog and Routing
Operating System:
ALL
Steps To Reproduce:
Hide

Create a two shard cluster. Assuming database "test" lives on "shard-rs1" initially, connect to mongos, and then:

// Create an unsharded collection mongos> db.mycoll.insert({a: 1}) // Move primary from the original home "shard-rs1" to "shard-rs0" mongos> db.adminCommand({movePrimary: 'test', to: 'shard-rs0'})

Now if you connect to shard-rs1 directly, you'll see that "test.mycoll" still exists on it.
Show
Create a two shard cluster. Assuming database "test" lives on "shard-rs1" initially, connect to mongos, and then: // Create an unsharded collection mongos> db.mycoll.insert({a: 1}) // Move primary from the original home "shard-rs1" to "shard-rs0" mongos> db.adminCommand({movePrimary: 'test' , to: 'shard-rs0' }) Now if you connect to shard-rs1 directly, you'll see that "test.mycoll" still exists on it.
Sprint:
CAR Team 2024-03-18
Linked BF Score:
129

When the feature flag gTrackUnshardedCollectionsUponCreation is enabled, on movePrimary the call to get unsplittable collections catalogClient->getUnsplittableCollectionNamespacesForDb() returns all unsharded collections.

However later on when we copy the DB over we create all the collections, but if the collection is a tracked collection we neither copy the indexes over nor copy the contents of the collection over either.

Seen on base commit 64dca7eaed26da5ee7547e1d62be599a87678327