Ghost sessions left on a server when using Pidgin client

Description

another thread: https://community.igniterealtime.org/message/239911#239911

Users report that when using Pidgin client they see that many sessions are left on the server after a client reconnects (after a sleep or hibernate). In this case a new session is generated for the client, but the old one is left and the server seems to send messages to that ghost sessions, so the messages are being lost. One user reports, that this can be eliminated by setting a resource in Pidgin manually (by default it generates a random string for every login it seems). This probably explains why those sessions are not being kicked out by the resource policy. But the server should still clear the stale sessions automatically. Users report that using 3.9.1 eliminates this issue. This may also be related to a memory leak reported in 3.9.2+ versions as stale sessions draw the memory.

Environment

Pidgin client

Activity

Show:

Michael Bear September 16, 2016 at 2:04 PM
Edited

Thanks to everyone who worked to report and test the software.
Happy to see the ghost session bug reported as resolved ... it sure wasn't any fun reverting version upgrades along the way (attempts from June 2014). winking face
And then an oops in packaging for 4.0.3 bit me until I got the fixed package (bad me for snagging the package well prior to the maintenance date!). slightly smiling face

I didn't make the time to upgrade to 4.0.2 before 4.0.3 was released ... but all appears to be well with 4.0.3. grinning face with big eyes

Thanks for the effort and software.

Kevin Stange April 1, 2016 at 3:15 PM

I have not seen this issue manifest since we upgraded to 4.0.2. Thank you for fixing this!

Daryl Herzmann March 21, 2016 at 4:07 PM

PR has been merged and Igniterealtime's Openfire hopefully verifies that this issue is resolved.

Daryl Herzmann March 16, 2016 at 9:01 PM

Igniterealtime is now running master build with the patch for https://igniterealtime.atlassian.net/browse/OF-464#icft=OF-464 mostly backed out. Will be watching closely for results.

Daryl Herzmann March 16, 2016 at 8:40 PM
Edited

I wonder if this commit is somehow to blame. We appear to be reproducing this with current master build on igniterealtime. I am going to revert that commit and try that build on igniterealtime to see what happens

Fixed

Details

Assignee

Reporter

Fix versions

Affects versions

Priority

Created June 25, 2014 at 12:35 PM
Updated September 16, 2016 at 6:49 PM
Resolved March 21, 2016 at 4:07 PM

Flag notifications