[pulsar-client-go] branch master updated: [fix] Fix data race in checkAndCleanIdleConnections (#981)

[zike]

This is an automated email from the ASF dual-hosted git repository.

zike pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/pulsar-client-go.git


The following commit(s) were added to refs/heads/master by this push:
     new 96bc58f  [fix] Fix data race in checkAndCleanIdleConnections (#981)
96bc58f is described below

commit 96bc58fd5e7ffb041703a082f81052a188f9c8ba
Author: Zike Yang <z...@apache.org>
AuthorDate: Tue Mar 7 17:15:00 2023 +0800

    [fix] Fix data race in checkAndCleanIdleConnections (#981)
    
    ### Motivation
    
    From the data race log of the issue:
    ```
    2023-03-07T02:05:50.0096902Z WARNING: DATA RACE
    2023-03-07T02:05:50.0097176Z Read at 0x00c0003bb298 by goroutine 111:
    2023-03-07T02:05:50.0097767Z   
github.com/apache/pulsar-client-go/pulsar/internal.(*connectionPool).checkAndCleanIdleConnections()
    2023-03-07T02:05:50.0098468Z       
/pulsar/pulsar-client-go/pulsar/internal/connection_pool.go:157 +0x24d
    2023-03-07T02:05:50.0099058Z   
github.com/apache/pulsar-client-go/pulsar/internal.StartCleanConnectionsTask.func1()
    2023-03-07T02:05:50.0099727Z       
/pulsar/pulsar-client-go/pulsar/internal/helper.go:25 +0x47
    2023-03-07T02:05:50.0099934Z
    2023-03-07T02:05:50.0100066Z Previous write at 0x00c0003bb298 by goroutine 
69:
    2023-03-07T02:05:50.0100537Z   
github.com/apache/pulsar-client-go/pulsar/internal.(*connection).connect()
    2023-03-07T02:05:50.0101126Z       
/pulsar/pulsar-client-go/pulsar/internal/connection.go:277 +0x5d2
    2023-03-07T02:05:50.0101649Z   
github.com/apache/pulsar-client-go/pulsar/internal.(*connection).start.func1()
    2023-03-07T02:05:50.0102222Z       
/pulsar/pulsar-client-go/pulsar/internal/connection.go:230 +0x34
    
    ```
    
    We can see that the read access of c.log here:
    
https://github.com/apache/pulsar-client-go/blob/c5956779f078176b6f915892dd47d9cbdb0d0fea/pulsar/internal/connection_pool.go#L157
    
    is conflict with the write access of c.log here:
    
    
https://github.com/apache/pulsar-client-go/blob/c5956779f078176b6f915892dd47d9cbdb0d0fea/pulsar/internal/connection.go#L277
    
    This caused the data race.
    
    ### Modifications
    
    * Using p.log to avoid accessing the c.log
---
 pulsar/internal/connection_pool.go | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/pulsar/internal/connection_pool.go 
b/pulsar/internal/connection_pool.go
index bf1e297..6ff7991 100644
--- a/pulsar/internal/connection_pool.go
+++ b/pulsar/internal/connection_pool.go
@@ -154,7 +154,8 @@ func (p *connectionPool) 
checkAndCleanIdleConnections(maxIdleTime time.Duration)
                        p.Lock()
                        for k, c := range p.connections {
                                if c.CheckIdle(maxIdleTime) {
-                                       c.log.Debugf("Closed connection due to 
inactivity.")
+                                       p.log.Debugf("Closed connection from 
pool due to inactivity. logical_addr=%+v physical_addr=%+v",
+                                               c.logicalAddr, c.physicalAddr)
                                        delete(p.connections, k)
                                        c.Close()
                                }