Describe the bug
In the last two days, I have found issues with boost that will have a ripple effect. The problem is that the set_lock() MySQL function can essentially block graphing and polling if a data source has been locked, and it is unable to be unlocked.
This will not happen on all occasions as this is a timing issue. As such Cacti needs to be a bit more intelligent about it's use of set_lock() in MySQL.
To Reproduce
Steps to reproduce the behavior:
-
Start boost
-
Abruptly cancel boost child process through a kill -9 which does not allow the signal handler to clear the lock.
-
Watch as boost basically gets hung up which can then lead to missed boost processes.
Expected behavior
Less issues that cause Cacti processes to essentially hang.
Describe the bug
In the last two days, I have found issues with boost that will have a ripple effect. The problem is that the
set_lock()MySQL function can essentially block graphing and polling if a data source has been locked, and it is unable to be unlocked.This will not happen on all occasions as this is a timing issue. As such Cacti needs to be a bit more intelligent about it's use of
set_lock()in MySQL.To Reproduce
Steps to reproduce the behavior:
Start boost
Abruptly cancel boost child process through a
kill -9which does not allow the signal handler to clear the lock.Watch as boost basically gets hung up which can then lead to missed boost processes.
Expected behavior
Less issues that cause Cacti processes to essentially hang.