High cpu-usage by system_server (AAS)

REPRODUCIBILITY: 100%
OS VERSION: 5.2.0.17
HARDWARE: 1.0.0.17
UI LANGUAGE: NL
REGRESSION: ?

DESCRIPTION:

Some time after starting AAS cpu-usage grows to 100% on one core.

PRECONDITIONS:

Having AAS enabled.

STEPS TO REPRODUCE:

  1. Enable AAS

  2. Wait

EXPECTED RESULT:

To be able to use Android-apps without my battery being drained in no-time.

ACTUAL RESULT:

Battery running empty quite quickly.

MODIFICATIONS:

Developer mode enabled.
Sshd enabled.
Community stores installed (Storeman, Chum).
Patchmanager installed (no patches installed).
Android AppSupport installed.
No audio- or WLAN-related system modifications.

Installed Android-apps:

  • ABNAmro
  • ChromiumBrowser
  • Bol.com
  • DigiD
  • FeedMe
  • Knab
  • Cardvault
  • Komoot
  • Morphe
  • Nine Email
  • Q-Park
  • Signal
  • Squeezer
  • Stratum
  • TomTom Go
  • WhatsApp

ADDITIONAL INFORMATION:

Stopping all Android-apps does not bring cpu-usage down.
Stopping AAS also does ~~not~ bring it down.
Restarting the phone is the only way (I found).
More information can be found here (and onward): Battery usage on JP26 - #36 by ID5

The thread that is using 100% is ‘binder:268_5’. Before my reboot it was also a binder-thread, but the number was different.

5 Likes

For easier reproduction it might be helpful to list your installed Android-apps.
Probably also a screenshot of which exact thread of system_server is the power hog.

And when exactly the CPU load jumps.

2 Likes

Hi there,

I can confirm the battery drainage described by @ID5 caused by <system_server> and further led to this bug report by @MrHappy . I did not have the problem before. The only thing I can remember I had done when it started to occur was: (maybe coicidential but these are the only things I can remember)

  • I tried the weather app “Cierzo”
  • allowed Whatsapp background process in system settings

After stopping AAS and restarting it <system_server> behaved normal again.
A reboot did not seem to work (start of AAS at boot was on)

After some time, I cannot tell how much, <system_server> is too high again. (phone gets considerably warm. (see picture upload)

1 Like

Let’s also post my list of android apps, although I don’t see much, or any overlap at all.

I cannot be 100% sure (I may have forgotten 1 or 2 apps, but those were already uninstalled), but let’s post the list in two parts, one before I noticed the problem and one list with apps after I noticed the problem, I also uninstalled some apps but the problem persists:

  • Aurora store
  • F-droid
  • microG settings
  • ING (Netherlands)
  • MijnKPN
  • Jottacloud
  • KNMI
  • Bitwarden
  • Firefox
  • Aegis
  • Discord (uninstalled, but already reinstalled as it did not change anything)
  • Thunderbird (Uninstalled)
  • K-9 Mail (uninstalled)

The following apps have been installed before I noticed the problem, but after that, I noticed the problem and made my first report on this forum:

  • foobar2000
  • Pano Scrobbler (uninstall did not help, currently installed again)
  • OHealth (currently uninstalled, gave up to try and pair my smart watch after I noticed the reported problem)

I also have installed a number of Android apps to try and find out what is wrong, but those have all been removed.

These are my Android apps:

Arcanechat
Whatsapp
Firefox clear
Öbb tickets
George
SchoolFox
Teams
Untis Mobile
Aurora Store
F-Droid
App Gallery
Lidl Plus
Easypark
Kelec
Meteor (not working by the way)
Klara (not working by the way)
Real Calc
MicroG Settings

There are Android apps that remain running in the background regardless of their privileges in SFOS and MicroG and Android extra settings.

Of my installed apps Yahoo Finance app is one that starts itself. It disappears when closed from SFOS UX, but after a minute or two it is back in the process list.

AFAIK any other similarly operating Android app can cause system_server load

Particularly any Instant Messaging apps are likely to cause load, but it is not limited to those.

Are you sure that you have set ‘Stop background activity on device closed’ (retranslation from my language) for such app?

There are a lot of system_server-processes running.
Unfortunately I do not know how to figure out what causes one of those processes to spin out of control.

SysMetrics will tell you more in this case then top.

SysMetrics can tell which thread (until now I have only seen ‘binder’) uses 100%.
That doesn’t provide (me) much more info. I would like to know what’s behind the binder-process (what app/service/…) and - better yet - have it fixed :slight_smile:

3 Likes

Indeed, same for me, a binder (+random number) is all that is shown that uses 100%.

SysMetrics tells me the same as ‘top -H’:
image
If you want something else/more, please tell me where/how to get that info :wink:

Can you narrow the hogging down to the exact PID?

If so: ps | grep <PID> might tell you more.

1 Like

I can narrow it down to the PID, but it gives me no more information:

39156 appsuppo {main} zygote64

Right now I witnessed this after stopping and restarting AAS due to the missbehaviour of system_server: (but it does not mean that WA has something to do with it, just reporting)

Well, haven’t seen such a drain before. I also notice that sometimes it takes a while to start (30+ minutes) or within a few minutes. Most of the time it is 99% for a binder thread and 50% for Jit Thread Pool. Not sure what where all active threads for this screenshot.

Another time I also noticed some other processes while AAS is running, while I had not yet startef any apps

1 Like

For me it sometimes seems that some kind of activity around receiving messages on WA or opening WA (or maybe something similar) triggers system_server. (aside: for me it’s only 13% of CPU that drains battery, Stopping and restarting AAS helps for some time)

1 Like

I noticed today for the first time that I’m also affected by this. I did notice that something was sometimes draining my JP2’s battery, but could not figure out what it was until now.

Just updated to latest .20 Finlayson OS version, this battery drain is still there. I also see some variant of “binder:XXX_X” as the culprit thread of the abnormal CPU load.

I’m not entirely sure this is exactly the same bug as OP reports, because in my case shutting off AAS also removes the CPU load, but this is a really bothersome bug, I started keeping AAS off when I’m not actively using it, but this is not ideal for my current needs.

How can I help troubleshoot this? Should this be brought up to Jolla during the next community meeting?

1 Like

I notice the same behavior on my phone. I could not check if that happened also on 5.2.0.18, the phone was discharging also quite fast. But with 5.2.0.20, I do see the same CPU load (~103%) by system_server.