power management should be disabled during ext4 migration |
||||||||||||||
Issue descriptionChrome Version: 61.0.3163.38 Chrome OS Version: 9765.21.0 Chrome OS Platform: samus Steps To Reproduce: 1. Update device to 61 so that container changes from M to N and ext4 migration is needed. 2. With a user profile of significant size (should take longer than 10 minutes), begin ext4 "OS update". Expected Result: Screen will remain on during OS update showing progress bar. Reboot will occur afterwards. Actual Result: Screen turns off after 7-8 minutes (seems to match default user power management settings). How frequently does this problem reproduce? (Always, sometimes, hard to reproduce?) 2/2 - Have not tested this on Chrome OS 60 boards to see if behavior is same. What is the impact to the user, and is there a workaround? If so, what is it? Can lead to user confusion during the update process. OS update screen claims that "computer will reboot when complete". When screen turns off due to power management, user may be confused. Did update finish? Fail?
,
Aug 17 2017
,
Aug 31 2017
bhthompson@ do you have any owner in mind for this issue?
,
Aug 31 2017
derat, is this something in your jurisdiction? dspaid, or is this something you would be able to help with? Either way this is not a regression, we did migrations in 60 without this, and this should not block 61.
,
Aug 31 2017
,
Aug 31 2017
This was discussed at length during the design of the migration process. https://docs.google.com/document/d/1-8hcSyTbnFh_Kzat4x2-K7J9zcg8JUG9rjTPBTtlVMY/edit#heading=h.nfkk2fmjp5xi https://docs.google.com/document/d/1csXazWS9UvZ4fdWN2jz_kbDKnF_cdOdPYrheemtijv8/edit#heading=h.hpn72ii2h7qu etc. The claim at the time was that there is no risk of data loss due to power loss. If you're just concerned about keeping the screen on, there are plenty of ways for Chrome to request that already. See device::PowerSaveBlocker with kPowerSaveBlockPreventDisplaySleep, for instance.
,
Aug 31 2017
Correct, there should be no risk of data loss due to power loss. We do still try to avoid mid-migration shutdown because its not feasible to test all corner cases and it provides a worse user experience. Back to the main topic though, I don't think keeping the screen on was ever discussed. I have no strong opinions one way or the other and will defer to PM/UX as to what the behavior should be. The actual logic to prevent screen-off should be pretty simple. We already have an "encryption_migration_active_" flag in PowerPolicyController to prevent shutdown on lid close, so it should just be a matter of also overriding the screen_off_ms fields. https://cs.chromium.org/chromium/src/chromeos/dbus/power_policy_controller.cc?dr=CSs&l=379
,
Sep 1 2017
So, it's not an issue of data loss, but it is more of how much do you expect your migration to happen. If it goes to sleep after 7 minutes and is waiting for user operation to resume from sleep, that's bad. Are we saying that the display goes off or are we saying that the machine goes to sleep?
,
Sep 1 2017
,
Sep 1 2017
The display goes off, the machine stays awake and finishes the migration. The concern in the initial report is just that it could be confusing to the user (they see the screen go off and assumed that the migration finished and rebooted).
,
Sep 1 2017
The intended behavior during migration is: - Prevent the system from sleeping - Do not prevent the display from going to sleep Did the system go sleep? If so, it sounds a regression.
,
Nov 7 2017
,
Nov 7 2017
did we come to a conclusion?
,
Nov 8 2017
We don't have a conclusion. The device won't sleep during migration. Only the screen will be turned. It is expected under the current implementation. IIUC, the suggestion in #1 is that keeping the screen on during migration should avoid user's confusion. I don't have a strong opinions on this. Can someone from PM/UX take this issue and decide the expected behavior?
,
Nov 8 2017
My personal opinion is that we should keep the display on, but I'll check on who the right PM/UX owner for this should be.
,
Nov 8 2017
+shihuis@ - Could you take a look and advise?
,
Nov 8 2017
If we wait long enough, the problem will take care of itself due to all users having already gone through the migration. :-P
,
Nov 11 2017
Long enough meaning months until we forcibly migrate users? ;) Let's keep the screen on during migration. This will ease user worry of what is happening by always showing the screen and progress. There will also be no misunderstanding of turning off vs screen falling asleep.
,
Nov 12 2017
Let's fill in some background information so that shihuis can make an informed decision: we're tralking about display going dark after a while; current 95%tile time to migrate is around 15 minutes. For those kind of users after 8 minutes screen goes blank but migration keeps on happening. After 8 more minutes the machine will reboot. Do we need to keep the screen on? (I guess maybe yes). possible argument against turning on the screen would be wasting battery.
,
Nov 14 2017
Thanks Junichi for the background. The use case I'm particularly worried about is users looking at their chromebooks and seeing that it's black, so thinking that the migration is done. The battery draining is a possible issue but given that the data is safe, always leaving the screen on should be more instructive.
,
Dec 8 2017
,
Aug 1
,
Aug 2
|
||||||||||||||
►
Sign in to add a comment |
||||||||||||||
Comment 1 by roy...@google.com
, Aug 17 2017