Bug #417
closedsessiond hangs when a session is stopped and other sessions are in progress
0%
Description
When several cluster-wide sessions are active and one of them is deactivated, sessiond seems to hang.
This causes a timeout in our application (TraceP), which happens 70 seconds after the call to lttng to stop the session.
The scenario is as follows:
- 4-node cluster each with a sessiond/consumerd pair
- 4 test apps running before the sessions are created. They are never stopped
- 4 cluster-wide sessions are created with several events and allowed to run for 10 sec.
- The sessions are stopped one after the other.
A log was collected with verbose mode on one of the nodes (SC-1). The issue takes place in several of the nodes, but it is seen for the 1st time on node SC-1 at the enf of the log (the rest of the log was omitted since it was a bit large)
Files
Updated by David Goulet about 12 years ago
- Status changed from New to Invalid
This is not a hang but a long stop command. There is possibly buffer bloating problem with the tc setup.
Closing the issue here to invalid since it does not seem to affect the core code or functionality of lttng-tools.