[ https://issues.apache.org/jira/browse/TS-4750?focusedWorklogId=26678&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-26678 ]
ASF GitHub Bot logged work on TS-4750: -------------------------------------- Author: ASF GitHub Bot Created on: 19/Aug/16 22:35 Start Date: 19/Aug/16 22:35 Worklog Time Spent: 10m Work Description: Github user shinrich commented on a diff in the pull request: https://github.com/apache/trafficserver/pull/863#discussion_r75561036 --- Diff: iocore/net/I_NetVConnection.h --- @@ -482,6 +482,7 @@ class NetVConnection : public VConnection /** Returns remote sockaddr storage. */ sockaddr const *get_remote_addr(); + IpEndpoint const &get_remote_endpoint(); --- End diff -- Cannot make it const because internally these functions cache the remote address and set a bit indicating whether the value is already cache or not. Issue Time Tracking ------------------- Worklog Id: (was: 26678) Time Spent: 1h 40m (was: 1.5h) > Erroneous WARNING: Connection leak from http keep-alive system > -------------------------------------------------------------- > > Key: TS-4750 > URL: https://issues.apache.org/jira/browse/TS-4750 > Project: Traffic Server > Issue Type: Bug > Components: Core > Reporter: Susan Hinrichs > Assignee: Susan Hinrichs > Fix For: 7.0.0 > > Time Spent: 1h 40m > Remaining Estimate: 0h > > We saw this a while back, but didn't get the fix pushed to open source. It > looks like the issue is still present in the current master. > HttpSessionManager caches the server address, but that cached address drifts > from the get_remote_addr() in the vc associated with the cached server > session. The problem is that one value is used to put the session into the > hash table, but the other value is used to remove the session from the hash > table later. So the session gets lost in the hash table. The session is not > found and the connection leak warning message is generated. -- This message was sent by Atlassian JIRA (v6.3.4#6332)