httpok restarts slow process

[hvelarde]

I have a Plone site with 3 Zope busy instances running under Supervisord; I have memmon and httpok plugins configured but seems that sometimes httpok restarts an instance that was recently restarted by memmon.

here is a piece of my memmon stderr log:

Checking groups app=1610612736
RSS of app:instance1 is 1614974976
Restarting app:instance1           <-- restarting
RSS of app:instance2 is 1297457152
RSS of app:instance3 is 1477554176
Checking groups app=1610612736
RSS of app:instance1 is 318668800  <-- going up
RSS of app:instance2 is 1297506304
RSS of app:instance3 is 1477554176
Checking groups app=1610612736
RSS of app:instance1 is 164720640  <-- down
RSS of app:instance2 is 1297575936
RSS of app:instance3 is 1477672960
Checking groups app=1610612736
RSS of app:instance1 is 340303872  <-- going up
RSS of app:instance2 is 1305280512
RSS of app:instance3 is 1477713920
Checking groups app=1610612736
RSS of app:instance1 is 166830080  <-- down
RSS of app:instance2 is 1318711296
RSS of app:instance3 is 1477849088
Checking groups app=1610612736
RSS of app:instance1 is 337248256  <-- going up
RSS of app:instance2 is 1325903872
RSS of app:instance3 is 1477685248
Checking groups app=1610612736
RSS of app:instance1 is 432963584  <-- stabilized
RSS of app:instance2 is 1325481984
RSS of app:instance3 is 1477685248
Checking groups app=1610612736
RSS of app:instance1 is 630874112
RSS of app:instance2 is 1325424640
RSS of app:instance3 is 1477685248

as you can see, instance1 RSS goes from around 300MB to 150MB a couple of times before stabilizing; this seems to me as an indicator of httpok restarting the instance in the middle. both plugins are configured running at TICK_60.

I solved this on the initial start by waiting 10 minutes before start using httpok; probably we need another parameter to deal with that after the process is running.

[mnaberez]

I solved this on the initial start by waiting 10 minutes before start using httpok; probably we need another parameter to deal with that after the process is running.

I'm not sure what this would look like but we'd be happy to review a PR.