+tasak@

The tab suspension happens when the tab is backgrounded for >30 mins. If the hang happens in 10 mins, it shouldn't be related to the tab suspension.

Also note that the tab suspension is enabled only on a very limited number of users. It's unlikely that you're hitting the Finch experiment.

I've verified that the tab seems to hang within about ~5 minutes of being
(re-)launched, so definitely not tab suspend.

I do have some extensions installed, so I'll check whether they are
interfering, but either way it sounds like P&S / MC are off the hook :)

Thanks for the verification!

(In any case you help is mandatory to confirm that MC solves the memory/jank problems on Chrome OS :-)

Thank you for the follow-up! Anyway we'll put more efforts on stabilizing MC and purge+suspend.

(BTW I've bought a personal chromebook for testing so I should be able to repro issues that are already reported :)

bashi: I've confirmed that my Chromebox does indeed have the PurgeAndSuspend experiment enabled.

My recommendation would be to get some basic status surfaced somewhere (whether through UMA, Task Manager, or whatever), so that we can establish easily whether unresponsive tabs are the suspended ones, versus some other cause.

Wez: Thanks for the follow-up. I created a CL to expose memory coordinator's state to task manager (https://codereview.chromium.org/2593233003/). That said, "suspended" state is not working correctly now. As of https://codereview.chromium.org/2590073002/ renderer suspension is disabled and currently MC just purges memory in "suspended" state. We will discuss how to re-enable suspending later (probably next year).

FWIW I just updated to 57.0.2970.0 dev and this issue no longer repros for me.

It sounds like this supports the theory that renderer suspension is the culprit?

Yeah, let me close this as Fixed for now. (We've disabled renderer suspend and this doesn't need to be a release block bug)

I'll try to find the root cause on my machine with custom build. I'll file another bug if needed.