[SERVER-33238] Prevent WiredTiger read ticket count from going negative Created: 09/Feb/18  Updated: 29/Oct/23  Resolved: 09/Feb/18

Status: Closed
Project: Core Server
Component/s: Concurrency, Diagnostics
Affects Version/s: None
Fix Version/s: 3.4.13

Type: Task Priority: Major - P3
Reporter: Kelsey Schubert Assignee: Ramon Fernandez Marina
Resolution: Fixed Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Issue Links:
Depends
Related
related to SERVER-32875 Don't stall ftdc due to running out o... Closed
is related to SERVER-33233 Don't stall ftdc due to running out o... Closed
Backwards Compatibility: Fully Compatible
Participants:
Case:

 Description   
Issue Status as of Feb 09, 2018

ISSUE SUMMARY
MongoDB 3.4.12 contains a bug that may cause WiredTiger read ticket count to become negative.

ISSUE IMPACT
Users may observe erroneous metric reporting. In particular, read queue metrics, which are derived from read tickets, will appear much larger than they actually are.

The default limit of 128 available read tickets becomes effectively larger, thus allowing more readers into the system. Workloads which hit the maximum read ticket limits and rely on this mechanism for back pressure may suffer performance degradation.

AFFECTED VERSIONS
Only MongoDB 3.4.12 is affected, and only when using the WiredTiger storage engine.

FIXED VERSIONS
A fix is included in MongoDB 3.4.13.



 Comments   
Comment by Ramon Fernandez Marina [ 11/Feb/18 ]

aqueen, 3.4.13 is now available for download.

Comment by Kelsey Schubert [ 09/Feb/18 ]

Hi aqueen,

I've updated the ticket description to include more details about this issue and its severity. To reiterate the new description, I expect most users running on MongoDB 3.4.12 will only observe reporting irregularities that may cause false alarm. However, workloads which hit the maximum read ticket limits and rely on this mechanism for back pressure may suffer, since the server will not enforce this limit.

We expect MongoDB 3.4.13 to be released soon, so I'd suggest waiting for it regardless of whether or not it's a significant issue for your workload.

Kind regards,
Kelsey

Comment by Githook User [ 09/Feb/18 ]

Author:

{'email': 'ramon@mongodb.com', 'name': 'Ramon Fernandez'}

Message: SERVER-33238 Revert the change that kept ftdc from stalling with no tickets

This reverts commit 4fa890fa8b6ba3b438213efcfab07b588fd58e63.
Branch: v3.4
https://github.com/mongodb/mongo/commit/fbdef2ccc53e0fcc9afb570063633d992b2aae42

Comment by Aaron Queen [ 09/Feb/18 ]

I had planned on upgrading some production servers soon to 3.4.12 and I'm worried about a potential issue. What is the severity of this problem? Is this issue something that could cause issues in production in 3.4.12 and should I hold off on upgrading until 3.4.13 is released with the change reverted or the underlying issue resolved?

Generated at Thu Feb 08 04:32:46 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.