Bug#816276: systemd: shutdown fails if swap can't be deactivated

[Michael Biebl]

Am 29.02.2016 um 17:35 schrieb Christian Pernegger:
> 2016-02-29 17:20 GMT+01:00 Michael Biebl :
>> This looks like a duplicate of
>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805133
> 
> Indeed it does, I do apologise. In my defence, I did filter for "swap"
> in reportbug and it didn't show up (still doesn't, actually).

The reason is, that it's already archived.
I've merged the two bug reports.

Michael


-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?



signature.asc
Description: OpenPGP digital signature

Bug#816276: systemd: shutdown fails if swap can't be deactivated

[Christian Pernegger]

2016-02-29 17:20 GMT+01:00 Michael Biebl :
> This looks like a duplicate of
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805133

Indeed it does, I do apologise. In my defence, I did filter for "swap"
in reportbug and it didn't show up (still doesn't, actually).

Thanks,
Christian

Bug#816276: systemd: shutdown fails if swap can't be deactivated

[Michael Biebl]

Hi Christian

Am 29.02.2016 um 12:43 schrieb Christian Pernegger:
> Package: systemd
> Version: 215-17+deb8u3
> Severity: normal
> 
> Hello,
> 
> I'm running jessie inside a low-memory (k)vm, 384M of RAM plus
> swap. This has been tuned so that as much as possible is swapped out
> during normal operation but there's no thrashing, i.e. the working set
> easily fits into RAM.
> 
> Shutting down this vm will fail more often than not because systemd
> tries to disable swap very early in the process, before the main
> daemon (logitechmediaserver) has shut down. The shutdown process will
> then hang until the vm is force-killed by libvirt after about 5 mins
> or so.

This looks like a duplicate of
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=805133

I've pulled the fix into our jessie branch, so it is very likely part of
the next stable point release.
If you want to test the fix, the code is at
https://anonscm.debian.org/cgit/pkg-systemd/systemd.git/log/?h=jessie
specifically
https://anonscm.debian.org/cgit/pkg-systemd/systemd.git/commit/?h=jessie=c4793975137d5d522fee104b7dab94a79547effd


Michael


-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?



signature.asc
Description: OpenPGP digital signature

Bug#816276: systemd: shutdown fails if swap can't be deactivated

[Christian Pernegger]

Package: systemd
Version: 215-17+deb8u3
Severity: normal

Hello,

I'm running jessie inside a low-memory (k)vm, 384M of RAM plus
swap. This has been tuned so that as much as possible is swapped out
during normal operation but there's no thrashing, i.e. the working set
easily fits into RAM.

Shutting down this vm will fail more often than not because systemd
tries to disable swap very early in the process, before the main
daemon (logitechmediaserver) has shut down. The shutdown process will
then hang until the vm is force-killed by libvirt after about 5 mins
or so.

1) Why is swap deactivated first thing? IMHO it should stick around
until after all "normal" processes have exited / been killed. After
all, if swap can be deactivated in the running system it wasn't really
needed in the first place.

2) Why does a failure to deactivate swap result in the shutdown
process hanging indefinitely? Shouldn't systemd either retry or fail
gracefully (i.e., ignore the problem)?

(For this specific use case I'll also look into giving the vm more RAM
and having the host swap out more, but I do not know yet if parts of
vm memory can be efficiently swapped out host-side.)

Regards,
Christian Pernegger



# fstab entry for swap
/dev/vdb none swap sw 0 0


# log of failed shutdown

Feb 26 20:29:00 slimserver systemd-logind[410]: Power key pressed.
Feb 26 20:29:00 slimserver systemd-logind[410]: Powering Off...
Feb 26 20:29:00 slimserver systemd-logind[410]: System is powering down.
Feb 26 20:29:01 slimserver nullmailer[839]: Stopping mail-transfer-agent: 
nullmailer.
Feb 26 20:29:01 slimserver swapoff[838]: swapoff: /dev/vdb: swapoff failed: 
Nicht genügend Hauptspeicher verfügbar # that's "not enough memory [available]"
Feb 26 20:29:01 slimserver systemd[1]: dev-vdb.swap swap process exited, 
code=exited status=255
Feb 26 20:29:01 slimserver systemd[1]: Unit dev-vdb.swap entered failed state.
Feb 26 20:29:01 slimserver sshd[405]: Received signal 15; terminating.
Feb 26 20:29:01 slimserver kernel:  vdb: unknown partition table
Feb 26 20:29:02 slimserver logitechmediaserver[840]: Stopping Logitech Media 
Server.
Feb 26 20:29:02 slimserver hwclock[837]: hwclock from util-linux 2.25.2
Feb 26 20:29:02 slimserver hwclock[837]: Using the /dev interface to the clock.
Feb 26 20:29:02 slimserver hwclock[837]: Last drift adjustment done at 
1456410669 seconds after 1969
Feb 26 20:29:02 slimserver hwclock[837]: Last calibration done at 1456410669 
seconds after 1969
Feb 26 20:29:02 slimserver hwclock[837]: Hardware clock is on UTC time
Feb 26 20:29:02 slimserver hwclock[837]: Assuming hardware clock is kept in UTC 
time.
Feb 26 20:29:02 slimserver hwclock[837]: Waiting for clock tick...
Feb 26 20:29:02 slimserver hwclock[837]: ...got clock tick
Feb 26 20:29:02 slimserver hwclock[837]: Time read from Hardware Clock: 
2016/02/26 19:29:01
Feb 26 20:29:02 slimserver hwclock[837]: Hw clock time : 2016/02/26 19:29:01 = 
1456514941 seconds since 1969
Feb 26 20:29:02 slimserver hwclock[837]: 1456514942.50 is close enough to 
1456514942.50 (0.00 < 0.001000)
Feb 26 20:29:02 slimserver hwclock[837]: Set RTC to 1456514942 (1456514942 + 0; 
refsystime = 1456514942.00)
Feb 26 20:29:02 slimserver hwclock[837]: Setting Hardware Clock to 19:29:02 = 
1456514942 seconds since 1969
Feb 26 20:29:02 slimserver hwclock[837]: ioctl(RTC_SET_TIME) was successful.
Feb 26 20:29:02 slimserver hwclock[837]: Clock drifted 0.4 seconds in the past 
104273 seconds in spite of a drift factor of 0.520768 seconds/day.
Feb 26 20:29:02 slimserver hwclock[837]: Adjusting drift factor by 0.319724 
seconds/day
Feb 26 20:29:02 slimserver dhclient[895]: Killed old client process
Feb 26 20:29:03 slimserver networking[876]: Deconfiguring network 
interfaces...Killed old client process
Feb 26 20:29:04 slimserver dhclient[895]: Internet Systems Consortium DHCP 
Client 4.3.1
Feb 26 20:29:04 slimserver dhclient[895]: Copyright 2004-2014 Internet Systems 
Consortium.
Feb 26 20:29:04 slimserver dhclient[895]: All rights reserved.
Feb 26 20:29:04 slimserver dhclient[895]: For info, please visit 
https://www.isc.org/software/dhcp/
Feb 26 20:29:04 slimserver dhclient[895]:
Feb 26 20:29:04 slimserver dhclient[895]: Listening on 
LPF/eth0/52:54:00:21:14:bf
Feb 26 20:29:04 slimserver dhclient[895]: Sending on   
LPF/eth0/52:54:00:21:14:bf
Feb 26 20:29:04 slimserver dhclient[895]: Sending on   Socket/fallback
Feb 26 20:29:04 slimserver networking[876]: Internet Systems Consortium DHCP 
Client 4.3.1
Feb 26 20:29:04 slimserver networking[876]: Copyright 2004-2014 Internet 
Systems Consortium.
Feb 26 20:29:04 slimserver networking[876]: All rights reserved.
Feb 26 20:29:04 slimserver networking[876]: For info, please visit 
https://www.isc.org/software/dhcp/
Feb 26 20:29:04 slimserver networking[876]: Listening on 
LPF/eth0/52:54:00:21:14:bf
Feb 26 20:29:04 slimserver networking[876]: Sending on   
LPF/eth0/52:54:00:21:14:bf
Feb 26 20:29:04 slimserver networking[876]: Sending on