[SERVER-22050] Primary vote for stepdown is unreasonably slow. Created: 04/Jan/16  Updated: 08/Jan/16  Resolved: 08/Jan/16

Status: Closed
Project: Core Server
Component/s: Replication
Affects Version/s: 3.2.0
Fix Version/s: None

Type: Bug Priority: Major - P3
Reporter: Hoyt Ren Assignee: Kelsey Schubert
Resolution: Duplicate Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: File StepDown_log.7z    
Issue Links:
Duplicate
duplicates SERVER-11086 Election handoff to new primary, duri... Closed
Operating System: ALL
Participants:

 Description   

I setup a fresh v3.2 cluster with 3 mongos, 3 config(as a replica) and 9 mongod(each 3 for a shard). It's completely empty now with only auth info. When I issue stepDown on a primary of a shard, I have to wait about 2 seconds before see the new primary. It's much slower than v3.0. During this time, the shard isn't available. Since I issue the command manually on an empty replica set, I believe the new primary should be vote immediately(scheduled maintenance for example), it's different from detect a failure.



 Comments   
Comment by Hoyt Ren [ 08/Jan/16 ]

Hi Thomas,

Thanks for remind me that the feature haven't implemented yet. Expected to see it in a near feature, it could help a lot about upgrade server and so on.

Comment by Kelsey Schubert [ 08/Jan/16 ]

Hi Hoyt Ren,

This is expected behavior. Under the new replication protocol, by default, elections occur less rapidly. To read more about this new replication protocol, please see our release notes. Under the new replication protocol, the time limit for detecting when a replica set's primary is unreachable can be configured. Please consider tailoring electionTimeoutMillis to better fit your needs. Additionally, you can select which protocol version you would like to use on your replicaset by specifying it in the replica set configuration.

There is an improvement planned to trigger an election more quickly following a step-down command. Please watch SERVER-11086 for updates and feel free to vote for it.

Thank you,
Thomas

Comment by Hoyt Ren [ 05/Jan/16 ]

Attached the file StepDown_log.7z

Comment by Kelsey Schubert [ 05/Jan/16 ]

Hi Hoyt Ren,

Thank you for the report. Can you please upload the logs of the old primary stepping down and the new primary stepping up?

Kind regards,
Thomas

Generated at Thu Feb 08 03:59:15 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.