[SERVER-62682] PrimaryOnlyService Does Not Call _rebuildCV.notify_all() leading to calls to waitForConditionOrInterrupt not being triggered - MongoDB Jira

XML

Word

Printable

Details

Type: Bug
Status: Closed
Priority: Major - P3
Resolution: Fixed
Affects Version/s: None
Fix Version/s: 5.3.0, 5.2.1, 5.0.7
Component/s: Internal Code
Labels:
- sharding-nyc-subteam1

Backwards Compatibility:
Fully Compatible
Operating System:
ALL
Backport Requested:

v5.2, v5.0
Steps To Reproduce:

Hide

Te test included in the PR 'LookUpInstanceHangsForever' reproduces the issue.

Show
Te test included in the PR 'LookUpInstanceHangsForever' reproduces the issue.
Sprint:
Sharding 2022-01-24, Sharding 2022-02-07
Linked BF Score:
35
Story Points:
2

Description

Since PrimaryOnlyService does not adhere to the contract of notifying the users of the condition variable whenever the underlying _state variable changes, its possible for its users such as lookUpInstance to end up waiting forever on the condition or interrupt

For instance on step down, the function updates the _state to kPaused but does not notify the condition variable

Adding a _rebuildCV.notify_all() after every update to _state should fix this issue.

Attachments

Issue Links

is related to

SERVER-51650 Primary-Only Service's _rebuildCV should be notified even if stepdown happens quickly after stepup

In Code Review

related to

SERVER-62884 Simplify synchronization semantics of `PrimaryOnlyService`

Open

Activity

People

Assignee:: Luis Osta (Inactive)

Reporter:: Luis Osta (Inactive)

Participants:: Githook User, Luis Osta

Votes:: 0 Vote for this issue

Watchers:: 5 Start watching this issue

Dates

Created:: Jan 14 2022 11:22:55 PM UTC

Updated:: Apr 07 2022 11:33:27 PM UTC

Resolved:: Jan 26 2022 02:07:44 PM UTC