Client slowing shutdown on Ubuntu

bkyn

I noticed that PIA is hanging when I'm shutting down (or rebooting) Ubuntu.

With a verbose Splash, I can see :

"A stop job is running for session 1..." with a 1min30 timer.

If I exit PIA before shutdown, everything is fine.

Did anyone has a fix for that?

Can we exit PIA through a script, just at shutdown?

doaks

Hi bkyn,

I'm very sorry that you're running into this. For now, I'd recommend clicking (or right-clicking) on the PIA tray icon and pressing Exit before shutting down. However, this is an issue that we're aware of and should be addressing soon.

muroc

Also experiencing the same thing on both Mint 18 and Ubuntu 16.  I can't remember to exit PIA prior to shutting down the OS, and I don't want to wait 90 seconds for thew computer to shut off.  Therefore I plan to stay with Mint 17.3 and Ubuntu 14.04 until PIA resolves this problem.

pjnsmb

Annoying for me too ! on siduction and sparky linux installs...................

I generally remember to exit PIA first but -

You can reduce the shutdown time by :

altering  /etc/systemd/system.conf:

DefaultTimeoutStopSec=10s

or to  a time that suits you 

bkyn

According to @doaks, it will be fixed in v66 (the next release).

pjc123

I am on Linux Mint 18.3 and PIA version 78 on two different laptops and I am experiencing the same issue as the OP on both.  Was it ever fixed, or is it back like Arnold Schwarzenegger?

In my case it is not session 1, it is session c2 as follows:

A stop job is running for Session c2 of user <removed> (40s/1min30s)

Max-P

It was fixed, but it does seem like it has returned under some cases, but not always. The annoying part is that Chromium also does that occasionally for me, so it seems like it's that Chromium bug that makes its way to nwjs and finally the PIA app.

I checked and it's still tracked in our bug tracker.

pjc123

At least it hasn't been closed.  You are right, it doesn't happen every time. Also, I am running Chrome, not Chromium.

Max-P

Yeah, it doesn't matter which browser you are using. PIA uses nwjs for its GUI, which is basically the same as Electron: it's a slimmed down Chromium browser turned into an application runtime running a local web page. That's why I'm mentionning Chromium specifically: it's not your browser, but the one PIA app uses.


If you want to speed up the shutdown, as a temporary workaround you can press Ctrl+Alt+Delete 7 times quickly. It signals systemd to abort waiting and to continue shutting down.

I had a systemd service a while ago that also "fixed" it, but this time around it would be nice if we could address the root cause of it and ensure the app dies properly on shutdown.

martouf

set the offending process to shut itself down upon receipt of SIGUSR2 @Max-P then make sending the signal part of shutdown

Max-P

@martouf It's a bit more complicated than that, the app already should respond to SIGKILL already so adding an extra signal wouldn't really make much of a difference. I think what happens is that on shutdown systemd just shuts down the display manager which kills Xorg which probably ends up with Chromium processes from nwjs hung waiting for X11.

I'll have to dig deeper into this to really find out what happens but I don't quite have the time to do that now.

I just remembered the timeout is configurable in /etc/systemd/system.conf as well. I think DefaultTimeoutStopSec is the property that controls this, setting it to something like 5-10 seconds could help avoid it getting stuck too long.

martouf

okay. it's something i implemented some years ago in a similar situation. had the SIGUSR2 sent very early in the shutdown sequence to avoid getting wedged later. YMMV, of course.