Boot from S3 state failing

subzero79 · 20. Mai 2016

Depending on the "get back" condition pm-utils passes the first positional argument with value "thaw" (hibernate) and "resume" (suspend) to scripts located in this folder /etc/pm/sleep.d

Examples can be found here

https://wiki.archlinux.org/index.php/pm-utils

As usual don't forget the executable bit.

This is an example to attend the issue with the official realtek driver

Bash

#!/bin/sh


PATH=/sbin:/usr/sbin:/bin:/usr/bin


case "$1" in


    thaw|resume)


        systemctl stop networking.service


        modprobe -r r8168


        modprobe r8168


        systemctl start networking.service


        ;;


    *) exit $NA


        ;;


esac


exit 0

Alles anzeigen

tinh_x7 · 20. Mai 2016

Will this script work with state 4 (suspend to disk)?
Put this script in /etc/pm/sleep.d right?

chmod u+x scriptname

subzero79 · 20. Mai 2016

Yes but you have to change the service line stop/start to init.
thats the folder

tinh_x7 · 20. Mai 2016

I'll give it a try if my Intel NIC keep hanging again.
Thanks.

tinh_x7 · 21. Mai 2016

Do the lines need to be change like this?
Where do you get the modprobe r8168 from?

Bash

#!/bin/sh


PATH=/sbin:/usr/sbin:/bin:/usr/bin


case "$1" in


    thaw|resume)


        systemctl init. networking.service


        modprobe -r r8168


        modprobe r8168


        systemctl init. networking.service


        ;;


    *) exit $NA


        ;;


esac


exit 0

Alles anzeigen

subzero79 · 21. Mai 2016

R8168 is specific for Realtek, if you have another card you need to find out which kernel module the card uses with lspci -vv or dmesg

tinh_x7 · 21. Mai 2016

e1000e is my module.

Code

00:19.0 Ethernet controller: Intel Corporation Device 15a1
        Subsystem: ASUSTeK Computer Inc. Device 85c4
        Kernel driver in use: e1000e

Update: I put that script in the /etc/pm/sleep.d, but I still got the eth0 hanging upon cron job shutdown.
By the way, I don't see init is a valid parameter of systemctl.
https://www.freedesktop.org/so…ystemd/man/systemctl.html

I got unknow operation when I tried these parameters:

Code

abc@server:~$ sudo systemctl init. networking.service
Unknown operation init.
abc@server:~$ sudo systemctl init networking.service
Unknown operation init
abc@server:~$ sudo systemctl init network.service
Unknown operation init
abc@server:~$ sudo systemctl init. network.service
Unknown operation init.


systemctl status network.service
Failed to get D-Bus connection: No connection to service manager.
systemctl status networking.service
Failed to get D-Bus connection: No connection to service manager.
systemctl status NetworkManager
Failed to get D-Bus connection: No connection to service manager.

Alles anzeigen

tinh_x7 · 24. Mai 2016

Good news,
It finally sleep and wake up as the cron job scheduled after several trial and errors and help from you guys.

Here are the steps for the solutions:

1. Use this script:

Bash

#!/bin/sh
PATH=/sbin:/usr/sbin:/bin:/usr/bin
case "$1" in
    thaw|resume)
        systemctl init networking.service
        modprobe -r e1000e
        modprobe e1000e
        systemctl init networking.service
        ;;
    *) exit $NA
        ;;
esac
exit 0

Alles anzeigen

2. sudo ethtool -K eth0 gso off gro off tso off

3. Disable C-State in the BIOS.

4. Restart the system.

Eryan · 24. Mai 2016

Just wanted to report back about my findings with my server. After changing the cronjobs to hibernate instead of suspend, disabling the flash memory plugin and re-enabling swap I stopped having the random freezes that I was getting. Seems to me then that the problem might be related to the lack of swap. I have some ram sticks laying around. I'll ramp up the server ram to 4 GB and return to the system to the old configuration (swap off, flash memory plugin and suspend) to see the problem comes back.

On a side note, hibernating the server instead of suspending translates in a few bucks less in the electricity bill. I think I'll stick to hibernation instead of suspension.

tinh_x7 · 24. Mai 2016

If you use hibernation like me, you don't have to disable the flash memory plugin, just enabled the swap, and leave the other feature on.

By the way, did you use any additional steps like me?

Eryan · 24. Mai 2016

Zitat von tinh_x7

If you use hibernation like me, you don't have to disable the flash memory plugin, just enabled the swap, and leave the other feature on.

I thought that would be a solution, but that defeats the purpose of the flash memory plugin, isn't it?

tinh_x7 · 24. Mai 2016

Kinda of.
If your system runs out of RAM, then it would activate the swap or activate when it's hibernating.
The more RAM you have, the less swapping required.
I"m using 32GB DDR3, so I'm not concerning much.

The flash memory have two features, the r/w system, and the swap.
I think the first feature for R/W is more important than the swapping.

Code

cat /etc/fstab
# /etc/fstab: static file system information.
#
# Use 'blkid' to print the universally unique identifier for a
# device; this may be used with UUID= as a more robust way to name devices
# that works even if disks are added and removed. See fstab(5).
#
# <file system> <mount point>   <type>  <options>       <dump>  <pass>
# / was on /dev/sda1 during installation
UUID=dbe4c621-3b9e-4d2e-9af3-003837d19629 /               ext4 noatime,nodiratime,errors=remount-ro 0 1
# swap was on /dev/sda5 during installation
 UUID=5fbde80b-57c6-4f7d-8b25-0ae9636bbdbe none            swap    sw              0       0
/dev/sdb1       /media/usb0     auto    rw,user,noauto  0       0
# >>> [openmediavault]
UUID=54bf67db-da31-4c50-bb3c-27140944b223 /media/54bf67db-da31-4c50-bb3c-27140944b223 ext4 defaults,nofail,user_xattr,noexec,usrjquota=aquota.user,grpjquota=aquota.group,jqfmt=vfsv0,acl 0 2
# <<< [openmediavault]

Alles anzeigen

Eryan · 24. Mai 2016

Well, that explains a lot. I am running on 2 GB. Going to 4 will perhaps help with the issue.

tinh_x7 · 24. Mai 2016

RAM is cheap now a day except DDR4.
You can upgrade to 8GB or 16GB if your system runs multiple tasks.

The only way I think that we can prolong the life of the SSD for a very long time with the plugin is to use a dedicated hard drive for the swap if we choose to use suspend-to-disk state. Or use a different mode.

ryecoaaron · 24. Mai 2016

If your system never swaps, then enabling it won't hurt anything (kinda obvious I know). The flashmemory plugin really protects you from the data logging that happens. I do recommend more memory though.

You can also create a swap file and put it on a data drive.

Eryan · 24. Mai 2016

Zitat von ryecoaaron

The flashmemory plugin really protects you from the data logging that happens.

This....

That's what I was missing. In that case, I'll go forward with the ram upgrade. I am stuck with DDR2 (the system is old but fully functional), but managed to get a good deal for 4 GB. Once the sticks arrieve I'll roll back to the old configuration (with swap and flash memory plugin) and see what happens.

ryecoaaron · 24. Mai 2016

If you are really worried about the swap file, you could plug in a cheap usb stick and put a swap file on it. The memory should really help too.

Eryan · 24. Mai 2016

The swap file is actually not a problem, but it seems that with only 2 GB of ram and no swap the system freezes from time to time after waking up from suspension. The problem does not happen if I re-enable the swap and send the server to hibernate instead. If increasing the ram solves the issue, then I would assume that 2 GB were not enough to suspend the system and in some cases that led to a system freeze. I really don't care if the swap is on or not, as you said, if there is enough ram the swap will basically sit there and should do no harm.

tinh_x7 · 24. Mai 2016

@ryecoaaronnullThe data logging that you referring to is the first feature, right?
UUID=dbe4c621-3b9e-4d2e-9af3-003837d19629 / ext4 noatime,nodiratime,errors=remount-ro 0 1

ryecoaaron · 24. Mai 2016

First feature?

Just to explain a bit more about this... The flashmemory plugin mounts the heavy write directories (like data logging) as tmpfs (ram drive/writes to memory). The problem with this is that if you fill tmpfs (usually defaults to half of system memory), it wants to write to swap. While it might be difficult to fill 2 gb memory on a fileserver, it isn't difficult to fill 1gb of file space especially if you have lots of logs. If it can't write to swap, it might lock up or reboot.

Boot from S3 state failing

Jetzt mitmachen!

Tags