[SERVER-5644] Auto-updating shards collection for changes in replica set is broken for user relying on DNS. Created: 18/Apr/12  Updated: 08/Mar/13  Resolved: 12/Feb/13

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

Type: Bug Priority: Major - P3
Reporter: Kyle Banker Assignee: Randolph Tan
Resolution: Done Votes: 1
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Operating System: ALL
Participants:

 Description   

https://groups.google.com/forum/?fromgroups#!topic/mongodb-user/G7b3DibumE4



 Comments   
Comment by Randolph Tan [ 13/Feb/13 ]

Copy and pasted from the Google threads, answered by Marc:

Hello. I just wanted to follow up on this thread.

We believe that you have experienced this issue: https://jira.mongodb.org/browse/SERVER-5058

The issue arises when isMaster() is called on a removed member of a
replica set. The removed member responds to isMaster(), and the
sharding configuration gets updated erroneously with the removed
member.

The work around is to manually update the shards collection in the
config db. (which it appears that you have already done.) and
restart each member in the cluster.

We apologize for the inconvenience. This issue is slated to be fixed
in version 2.1.2.

Comment by Sebastian Dahlgren [ 13/Feb/13 ]

I noticed that this issue was resolved with "Community answered". What did the community answer?

Cheers

Comment by Sebastian Dahlgren [ 19/Apr/12 ]

During an upgrade from 2.0.1 to 2.0.4.

While updating mongod to 2.0.4, our config servers was still running 2.0.1.

Comment by Eliot Horowitz (Inactive) [ 19/Apr/12 ]

That seems like the real issue.
What version was this?

Comment by Sebastian Dahlgren [ 19/Apr/12 ]

@Eliot, we waited for a bit over 30 minutes, but the old configuration in the 'shards' collection did not come back. We ended up with only the node that we previously made a rs.remove() on.

Comment by Eliot Horowitz (Inactive) [ 19/Apr/12 ]

I don't get the issue.
Sounds like if they just waited a few more minutes it would have updated to the new config?

Comment by Kyle Banker [ 18/Apr/12 ]

Perhaps we need a setting to prevent the replica set from updating the config servers in this case?

Generated at Thu Feb 08 03:09:30 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.