changing IO Scheduler

informant

Active Member
Jan 31, 2012
728
8
38
Hi, i have a question.

If we change the IoScheduler permanently to cfq (http://pve.proxmox.com/wiki/IO_Scheduler), do we must only add this line
Code:
GRUB_CMDLINE_LINUX_DEFAULT="... elevator=cfq"
in /etc/default/grub and run update-grub
or must we change the "..." to a path or so?

please answer. thanks and regards
 

WhiteStarEOF

Active Member
Mar 6, 2012
96
10
28
The ... is just a place holder for any other options that may already be specified there. Usually the line is blank. So when you're done modifying the file, your line should look like this,

Code:
GRUB_CMDLINE_LINUX_DEFAULT="elevator=cfq"
 

informant

Active Member
Jan 31, 2012
728
8
38
hi whitestarof,

thanks for reply.

our grup is:
Code:
...
GRUB_CMDLINE_LINUX_DEFAULT="quiet"
GRUB_CMDLINE_LINUX=""
...

what we must change here, can you post it complete please. very thanks.

regards
 

WhiteStarEOF

Active Member
Mar 6, 2012
96
10
28
So in this case, you'd want it to look like this,

Code:
GRUB_CMDLINE_LINUX_DEFAULT="quiet elevator=cfq"

And I believe there is a command to get Grub to pick up the change, but I do not recall what that command is. After that, you should be able to reboot, login, and cat /sys/proc/sda/queue/scheduler. cfq should be selected.


edit:
Out of curiosity, are you making this change due to poor disk performance? That's exactly the reason I did it. I was going to start up a thread about IO schedulers, but haven't got to it yet. Could you tell us a little bit about the disk setup on your Proxmox install?
 
Last edited:

WhiteStarEOF

Active Member
Mar 6, 2012
96
10
28
Well be sure to report back on your results. If the change to Deadline is a problem, the devs won't know that until people start speaking up.

For me, the change back to cfq boosted disk performance quite noticeably. I never did get around to putting hard numbers behind it. But under deadline, VMs would take a considerable amount of time to boot, and would be rather sluggish in responding, even if no applications were running. Changing to cfq made the VMs much more responsive almost immediately. The key factor, changing back to deadline would create the sluggish performance issue again.

I'm using a Raid 5 with 3 disks presented to Proxmox via iSCSI.
 

informant

Active Member
Jan 31, 2012
728
8
38
@sumsum,

if deadline not work by very many users, we must use the stable scheduler cfq. if you have a solution for deadline scheduler, please tell us this :)

@whitestarof,

after change the backups run normal and without hight load. all users with the problem of io load must be change back to cfq. it´s works fine and without errors.

regards
 

mmenaz

Active Member
Jun 25, 2009
828
20
38
Northern east Italy
I'm surprised, because I experienced in the past (2.x) just the opposite, and CFQ was a real problem to me, i.e. in the past I've used configuration like:
- sata for Proxmox installation
- RAID10 sas for VM (lvm storage)
and I noticed that I/O on the sata was killing the I/O on the RAID10 too, due to the CFQ scheduler
Also I have sata for local backup and iso images, and transferring a big iso there impacted a lot the I/O on the RAID too.
In short, every I/O on slow hardware was putting in it's knee the whole system.
Have a look at this thread:
http://forum.proxmox.com/threads/10...sponsiveness-even-if-on-different-SAS-storage
so
a) deadline is working badly due to some other reason
b) everyone is wrong ;P
regards
 

informant

Active Member
Jan 31, 2012
728
8
38
hi mmenaz,

shure, but in version 2.* its works fine. in version 3 is only changed the scheduler to deadline. and with this comes the problem. if you say, it´s a other problem, than tell us, what is the problem, well many users in this forum have this problem with the io load/scheduler.

regards
 

informant

Active Member
Jan 31, 2012
728
8
38
@all and @WhiteStarEOF,

the problem with hight io is back with cfq scheduler. we have changed it without restart, but on 2nd auto-backup the io load goes hight. the bug is open soon.

io-load-on-backup.png

any ideas for fix?

regards
 
Last edited:

informant

Active Member
Jan 31, 2012
728
8
38
morning sumsum,

thanks for help and reply,

we use a hp proliant dl 360 g5 and other hp servers with smart array p400 (bbu + 512 mb cache) and on it 6x 1 tb seagate enterprise hdd in a raid 5 configuration.
as backupstorage we use a synology 1812+ with a raid 5 (hdds are western digital enterprise 3tb) in ext4 format -> in proxmox as nfs share. as info, the same storage are using in proxmox 2.* and other servers whitout this error. we think, it´s not the storage, well in v2 it´s works fine.

do you need more informations?

regards
 
Last edited:

Voleatech

New Member
Mar 22, 2013
25
0
1
Thanks for reporting back @informant. I still have the same issue you have on different hardware.
Running a dell T420 with a perc H710 on a raid one with 1TB western digital drives.
Snapshots worked fine until 3.0 and now I get high io loads.
I hope the proxmox team finally looks into this and is not discarding it saying it is not a bug
 

tom

Proxmox Staff Member
Staff member
Aug 29, 2006
15,558
927
163
99 % of backup problems can be fixed with configurations adaptions. If you think its a bug, the issue should be reproducible. Then we will debug it (I mean it works here).

If you issue started with 3.0, switch back to cfq (see also http://pve.proxmox.com/wiki/Roadmap#Proxmox_VE_3.0)
 
Last edited:

Voleatech

New Member
Mar 22, 2013
25
0
1
Hi Tom,
thanks for your answer.
I will try to reinstall Proxmox 3.0 from scratch as soon as I can get my production server of the net for a few hours, most likely next weekend. Maybe the upgrade affected the system somehow. I will report back if it works then.
 

tom

Proxmox Staff Member
Staff member
Aug 29, 2006
15,558
927
163
Just to note, for changing the IO scheduler a re-install is not necessary.
 

Voleatech

New Member
Mar 22, 2013
25
0
1
I know but I want to rule out that something went wrong with the upgrade.
If the error persists after the reinstall I can at least narrow it down a bit more.
 

informant

Active Member
Jan 31, 2012
728
8
38
hi tom,

we have changed the scheduler back to cfq and have the same issue on 2nd auto-backup. also the first backup run mostly (after a reboot) and the next have this issue. sporadical it´s the first backup. we can reproducible this error. what can we do or what do you need for information to help us, to find the error/bug?

please reply tom, very thanks.

regards
 
Last edited:

About

The Proxmox community has been around for many years and offers help and support for Proxmox VE, Proxmox Backup Server, and Proxmox Mail Gateway.
We think our community is one of the best thanks to people like you!

Get your subscription!

The Proxmox team works very hard to make sure you are running the best software and getting stable updates and security enhancements, as well as quick enterprise support. Tens of thousands of happy customers have a Proxmox subscription. Get your own in 60 seconds.

Buy now!