may be leaky

[Dieterbe]

i just saw start_graphite.py grow to >26GB, both virt and RSS.
this was with graphite-raintank and also modified to run in debug mode

#os.execv('/usr/local/bin/gunicorn', args)
from graphite_api.app import app

app.run(debug=True, port=8888, host="0.0.0.0", use_reloader=False)

and then finally

[11709.172545] Out of memory: Kill process 25436 (start-graphite.) score 852 or sacrifice child
[11709.172547] Killed process 25436 (start-graphite.) total-vm:28044608kB, anon-rss:27951128kB, file-rss:0kB
[11714.273750] mce: [Hardware Error]: Machine check events logged

so not sure what's going on here.

[Dieterbe]

@woodsaj anything actionable here? if not let's just close this

[woodsaj]

nothing that can be actioned here without knowing the query load that was being sent.

performance testing definitely cant be done when using Flask's built in HTTP server.