Limiting use of RAM in Chrome?
I wrote a Python 2.5 program which kills chrome's renderers when they use over a set amount of memory. I run this program under
watch
. (note that it uses the psutil module which isn't included with Python.)import sys, os, psutil if len(sys.argv) == 2: try: limit = int(sys.argv[1]) except: limit = 200 # default 200MB else: limit = 200 uid = os.getuid() for p in psutil.get_process_list(): try: if (p.name == 'chrome' and any('type=renderer' in part for part in p.cmdline) and p.uid == uid): m = p.get_memory_info() #print p.pid,m, m.rss / 1024 / 1024, m.vms / 1024 / 1024 if (m.rss / 1024 / 1024) > limit: # kill if rss is greater than limit print 'Killed', p.pid p.kill() except psutil.error.NoSuchProcess: pass except psutil.error.AccessDenied: pass
I rely on Session Buddy to recover the open tabs when chrome fails to restore them.
The only thing I've seen to date that can do this is to run chrome inside a container and limit the containers ram.
However this has some major caveats,
Running chrome is complicated by the dockerize setup and launch sequence
for one, Chrome already uses kernel containers to sandbox its threads; so you have to run the container with a kind of root privilege that allows that to work. This can be circumvented, and the linked container model does so. (it does practically everything it needs to)
You will almost certainly loose gpu acceleration
getting audio to work is complicated, but handled in the linked container model.
Whatever else you expect to go wrong when you void your warranty, Chrome violently dislikes being told not to use more ram, and will act up and tantrum accordingly.
But it ultimately does work.
I am more interested in applying these ram limits to Electron Shell apps which don't have prebuilt docker images to rangle them for you.
Off topic but I want to note that Firefox is very well behaved on limited hardware, but I don't consider that a real answer.
It isn't necessarily plugin. Note that webpages are no longer static. Some webpages just have a non-trivial amount of async activity going on. Add on the activity from the plug-ins and you got some unknowns.
The best remedy I have found is to kill the webpage and reload it. AFAIK, there is no way to limit the amount of RAM a webpage uses.