Red Hat Bugzilla – Bug 15911
pulse fails to correctly restart
Last modified: 2005-10-31 17:00:50 EST
Problem: According to the documentation, you should be able to run '/etc/rc.d/init.d/pulse restart'. Running it with the restart command fails to clean up the old pid file causing pulse to complainthat it is still running or that you need to make sure you are running the command as root even if you ARE root. It also leaves lvs and nanny running and does not kill them off. Workaround: run the command as '/etc/rc.d/init.d/pulse stop ; /etc/rc.d/init.d/pulse start' though this too does NOT kill lvs or nanny. You end up having to run killall -9 lvs; killall -9 nanny in between the stop and start commands to pulse. According to the docs, the accepted way to restart the LVS software is to run either pulse stop and start or use restart. (unless I'm reading this wrong
This strikes me as very likely being a real bug as I do not believe it was included in formal testing. I will investigate.
Accidentally flagged as resolved instead of accepted
Update: Because of LWE, no personnel are availabel to investigate this problem until after 8/21.
Additional information on the work-around. stop does indeed stop lvs. anny, and all other piranha components. It is possible that because the stop and starts are back-to-back, with no testing logic, that a delay is needed between the 2 commands. Also, performing a "kill -s SIGTERM xxx" to nanny or lvs would be better than -9, as this should cause all the children to terminate properly.
I have verified this problem, the issue is that depending on the number of virtual and real servers, the time to stop can be pretty long. (Therefore a simple sleep will not be a safe answer). I see that there is some actual reload code being worked on (As opposed to stop/start). -- Keith Moore
Understood. It will be looked at.
*** This bug has been marked as a duplicate of 57654 ***