[SERVER-13024] Clear the 'jumbo' flag from the chunk document when the chunk is no longer over-sized Created: 04/Mar/14  Updated: 01/Feb/24

Status: Blocked
Project: Core Server
Component/s: Sharding
Affects Version/s: None
Fix Version/s: None

Type: New Feature Priority: Major - P3
Reporter: William Zola Assignee: Backlog - Catalog and Routing
Resolution: Unresolved Votes: 1
Labels: car-product-sync, oldshardingemea
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Depends
depends on SERVER-86055 Automatically clear jumbo flags befor... Needs Scheduling
Duplicate
is duplicated by SERVER-36059 Can't move jumbo chunks even after ch... Closed
Related
related to SERVER-11629 Allow Jumbo Chunks to be Drained / Mi... Closed
related to SERVER-14121 Store size for chunk in config.chunks. Closed
related to DOCS-4163 Document how to manually reset Jumbo ... Closed
is related to DOCS-5701 Jumbo chunks are never cleared automa... Closed
is related to SERVER-13806 Need better detection and reporting o... Closed
is related to SERVER-21931 Only mark chunks as jumbo if split fa... Closed
is related to SERVER-40435 A clearJumboChunk command to clear th... Closed
Assigned Teams:
Catalog and Routing
Participants:

 Description   

Currently, if the balancer goes to split a chunk and cannot, the chunk gets marked as 'jumbo', which will prevent the balancer from migrating that chunk and from trying to split it again.

If in the future, documents are deleted from that shard key range or the cluster's chunksize setting changes, the chunk itself may no longer be 'jumbo', but MongoDB will not clear the 'jumbo' flag from that chunk.

If you have a system where documents are short-lived (say, one week) and are then removed from the system, this will result in a system with a large number of chunks that are incorrectly marked 'jumbo'. In addition, the number of 'jumbo' chunks will grow over time.



 Comments   
Comment by Kaloian Manassiev [ 06/Nov/21 ]

tommaso.tocci, I believe this might belong under the AutoSplitter project's subsequent milestones (if we will be doing something smarter around tracking deletes).

Comment by Kaloian Manassiev [ 27/Jul/18 ]

One thing that should also be noted is that since the balancer utilizes the catalog cache to get the jumbo status of chunks, it is not sufficient to just unset the jumbo field on the chunk, because due to incremental refresh these changes will not be seen in the balancer's cache.

Generated at Thu Feb 08 03:30:24 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.