[SERVER-8279] Warn or do not allow the balancer to balance when running in a mixed sharded cluster Created: 22/Jan/13  Updated: 07/Dec/23  Resolved: 11/Feb/13

Status: Closed
Project: Core Server
Component/s: Sharding
Affects Version/s: 2.0.8, 2.2.2
Fix Version/s: 2.4.0-rc1

Type: Improvement Priority: Major - P3
Reporter: Andre de Frere Assignee: Greg Studer
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: File mixed_mongos_v20_and_v22.js    
Issue Links:
Problem/Incident
causes SERVER-83956 Balancer wrongly emit warning message... Open
Related
related to SERVER-6902 MongoDB 2.2 and MongoDB 2.0 cannot be... Closed
Participants:
Linked BF Score: 0

 Description   

SERVER-6902 means that running 2.2 and 2.0 in a sharded environment is a very bad thing. Given this, it should be very apparent that this is happening - difficult to miss messages in the log would be a first step, potentially not allowing the balancer to operate would be the most intrusive step.



 Comments   
Comment by Greg Studer [ 07/Feb/13 ]

Think the new warning alongside the newer exclude functionality in https://github.com/mongodb/mongo/blob/master/jstests/multiVersion/config_version_excludes.js (not yet fully doc'd) will give more visibility to the problem in the future.

Comment by auto [ 07/Feb/13 ]

Author:

{u'date': u'2013-01-28T19:35:26Z', u'email': u'greg@10gen.com', u'name': u'Greg Studer'}

Message: SERVER-8279 warn if shard versions differ in mixed cluster
Branch: master
https://github.com/mongodb/mongo/commit/6dec70a4f6d8b968fe69af32ce0efc2198c13eed

Comment by Greg Studer [ 23/Jan/13 ]

Tried to reproduce this with v2.2.0 and v2.2.2 - no luck. SERVER-6002 was an issue running v2.0 and v2.2 mongoD processes together, not mongoS. It was intended that v2.0 and v2.2 shards should be compatible, however a bug prevented this (which was fixed).

Think a mixed-cluster warning is a good idea in general, but not yet able to see a mixed mongoS issue.

Comment by Jon Hoffman [ 22/Jan/13 ]

Hi, I observed this problem at Foursquare. Here are the details:

We were running a cluster where the mongoS versions were mixed between 2.2 and 2.0 versions (the mixed versioning was accidental). One of the 2.2 mongoS got the balancer lock and moved some chunks. The 2.0 mongoS did not update their chunk configuration when the move was complete. So all traffic continued to be directed to the old shards for those chunks.

Generated at Thu Feb 08 03:17:00 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.