Hi Proviewers,
Thank you to all those who support and contribute to ProviewR, it is really an amazing system and have been able to build a great feedmill control system using it running on Raspberry Pi 3B and 4B’s.
I have been experiencing a problem that appears to have got worse with version 5.7 since I upgraded from 5.6 in December. I experience the runtime graphs closing and freezing of the entire operating system.
The backup program rt_bck service also stops working and sometimes states the pwr_backup.bck backup file doesn’t exist.
This used to happen sometimes on version 5.6 as well, but I was developing and restarting the runtime often so didn’t notice it as much.
I ran the “top” command to monitor the processes and noticed rt_xtt_gtk kept growing in the amount of RAM used until the system RAM was exhausted and experienced the freezes and crashes.
At first though it was a problem with my code or the Raspberry Pi build on 4B so I setup a Proview system with the demo package on Ubuntu 18.04.
I experienced the same problem with the demo project ruling out the Raspberry Pi build as well as my code as the cause for the problem which appears to be a memory leak.
The 5.7 release only includes the GTK build so I’m not sure if its a rt_xtt problem or a problem in the GTK flavour.
The runtime only appears to increase if a Ge graph or Trend is opened and only having the runtime navigator open does not cause the RAM to increase.
As different graphs are opened the RAM increases but it never gets less even when you close the window.
Opening the same graph again does not increase the RAM as much as the first time but over time RAM usage still increases.
I am using 5.7 in a production environment on Raspberry Pi 3B and 4B with networked nodes on Raspbian Buster with only this problem.
Proview works very well expect for the memory leak. I have got around it by using an app call Monit which monitors the rt_xtt_gtk and closes and reopens it when it reaches 75%.
Since I have done this I get no more crashes and freezing or the problem with rt_bck. I suspect 5.6 was the same so see no point rolling back.
The Workbench does not show a memory leak even if you Preview the graph in the Ge so it appears to be only in the runtime.
Please can someone verify this and hopefully find the cause. I’m too busy with projects to dig deep into the code and have already spent weeks trying to diagnose it.
George