raid1 and hibernate

noobuser · 02-05-2010, 04:57 PM

Hi,I'm a new slackware user.
I'm trying to build my mediacenter xbmc with Slack 13.
I've an Amd system with 2 disks configured in raid1.
A / partition as md0, /home partition as md1 and swap as md2
It works everything fine, also suspend to ram (S3) ,but not hibernate(S4).
My problem is that during restart ,the kernel sees the swap partition (md2) "dirty" and starts immediately the background rebuild deleting my suspended image.
What I should do to fix this issue?Is there some kernel parameter to put in lilo.conf?
In lilo.conf I have these parameters:

Code:

boot = /dev/md0                                                                                        
raid-extra-boot=mbr-only                                                                               
compact        # faster, but won't work on all systems.                                                
# Bitmap in BMP format: 640x480x8                                                                      
  bitmap = /boot/slack.bmp                                                                             
# Menu colors (foreground, background, shadow, highlighted                                             
# foreground, highlighted background, highlighted shadow):                                             
bmp-colors = 255,0,255,0,255,0                                                                       
bmp-table = 60,6,1,16                                                                                
   
bmp-timer = 65,27,0,255                                                                               

# Append any additional kernel parameters:
append=" root=/dev/md0 vt.default_utf8=0 resume=/dev/md2"
prompt
timeout = 50
vga = 773
# vga = normal
# ramdisk = 0     # paranoia setting

image = /boot/vmlinuz-2.6.32.7-smp
  root = /dev/md0
  label = 2.6.32.7
  read-only

# Linux bootable partition config ends

If you need any other info ,just ask

Thank you

ivandi · 02-06-2010, 09:20 AM

You do not need to put the swap on raid device. Just add your two swap partitions to fstab and the kernel will use them in parallel.

noobuser · 02-07-2010, 04:30 AM

Quote:

Originally Posted by ivandi

You do not need to put the swap on raid device. Just add your two swap partitions to fstab and the kernel will use them in parallel.

You mean that the kernel will write the same data on BOTH the swap partitions?
I don't think.I'm looking for redundancy also on swap.If I hibernate my system on /dev/sda3 and the next time I resume my system,/dev/sda is dead,I'll lost my system image.

I've found the way anyway,using initrd.The only problem is that the resume is very slow.Probably there is still some recovery of /dev/md2 during resuming.Slackware's initrd script doesn't manage the md_mod.start_ro parameter.
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=415441

I need to understand how to add that option.

ivandi · 02-07-2010, 11:46 AM

Never tried it myself but may be you should use a swap-file.

filefrag -v /swapfile should return the resume_offset (step 2).

noobuser · 02-09-2010, 07:04 AM

Quote:

Originally Posted by ivandi

Never tried it myself but may be you should use a swap-file.

filefrag -v /swapfile should return the resume_offset (step 2).

Thanks for the hint.
I've tried,but I have some problem to manage correctly the resume paramters for the kernel.I think that initrd script doesn't manage correctly the parameters for swap over file.Infact the system at boot time fails looking for the swap file (UUID device not found),but I could be wrong.

Anyway now I'm able to resume correctly from swap device on partition /dev/md2,adding the parameter

Code:

echo 1 > /sys/module/md_mod/parameters/start_ro

in /init script of initrd image.The only problem is that it's really slow,more slow than cold boot.On a single disk configuration (no raid then) it works fine.
Do you have some idea about this problem?