[SERVER-24960] Shard may not realize that collection is already sharded after restart Created: 08/Jul/16  Updated: 06/Dec/22  Resolved: 10/Nov/17

Status: Closed
Project: Core Server
Component/s: Sharding
Affects Version/s: 2.6.4, 3.0.9, 3.3.9, 3.4.2, 3.5.4
Fix Version/s: None

Type: Bug Priority: Major - P3
Reporter: Randolph Tan Assignee: [DO NOT USE] Backlog - Sharding Team
Resolution: Duplicate Votes: 0
Labels: todo_in_code
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: File zero_version.js    
Issue Links:
Duplicate
duplicates SERVER-32198 Missing collection metadata on the sh... Closed
Related
related to SERVER-27308 Reduce cost of sharding check in Docu... Closed
is related to SERVER-24762 support views of sharded collections Closed
Assigned Teams:
Sharding
Operating System: ALL
Participants:

 Description   

Until a mongos who knows that the collection is sharded informs the shard. The issue is that shards assume that collections are unsharded by default and it should double check with the config servers if this is the first time it sees a namespace and mongos sends a version of (0, 0|0).



 Comments   
Comment by Kaloian Manassiev [ 10/Nov/17 ]

Failure to refresh will not fassert, it will just keep failing queries.

Comment by Esha Maharishi (Inactive) [ 10/Nov/17 ]

kaloian.manassiev hmm, yes I think so, since if a donor fails to refresh after a migration (issue described in SERVER-31056), it will fassert and have to be restarted, which will then cause the issue in this ticket.

Comment by Kaloian Manassiev [ 10/Nov/17 ]

esha.maharishi, can this be closed as duplicate of SERVER-31056?

Comment by Esha Maharishi (Inactive) [ 13/Mar/17 ]

Bringing this back into Needs Triage so we can decide whether this can be easily fixed in 3.6, or if it will be automatically fixed as part of either:

  • moving metadata commands to the config server
  • changes to migrations under transactions

The query team believes the "kickback" or extra round trip from shard to mongos will cause a significant enough slow down for queries on views for users to start complaining.

Comment by Geert Bosch [ 02/Aug/16 ]

Removing from non-materialized views, as we worked around this issue by always treating everything as sharded.

Generated at Thu Feb 08 04:07:53 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.