Bug 251617

Summary: pm-utils sets the VM background timeout to 300 milliseconds!
Product: [Fedora] Fedora Reporter: Arjan van de Ven <arjan>
Component: pm-utilsAssignee: Phil Knirsch <pknirsch>
Status: CLOSED CURRENTRELEASE QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: low Docs Contact:
Priority: low    
Version: 8CC: bnocera, mgarski, mozilla_bugs, opensource, rvokal
Target Milestone: ---   
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: pm-utils-0.99.4-10 Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2008-05-18 05:21:31 EDT Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---
Bug Depends On:    
Bug Blocks: 204948, 418441    

Description Arjan van de Ven 2007-08-09 23:33:18 EDT
Description of problem:

the laptop-tools script sets the VM background write timeout to 30 milliseconds;
this is INSANE. It keeps the disk extremely busy, waking it up all the time,
negating any power savings that the VM caches give you, and it also really
really sucks for peformance.
Comment 1 Osma Ahvenlampi 2007-08-10 19:48:39 EDT
Actually, no, it sets the timeout to 0.3 seconds, which is rather reasonable in
terms of "this data should be written to disk" while the computer is on AC power.

>       # Set dirty page values
>       echo $DIRTY_WRITEBACK > /proc/sys/vm/dirty_writeback_centisecs
>       echo $DIRTY_EXPIRE > /proc/sys/vm/dirty_expire_centisecs
...
>               DIRTY_WRITEBACK=30
>               DIRTY_EXPIRE=30

At least, that's what it does on an up-to-date Fedora 7.
Comment 2 Arjan van de Ven 2007-08-10 23:10:19 EDT
you're right about the 300msec part...

but I strongly disagree about this being appropriate or reasonable....
The kernel default is 5 seconds and that is on the conservative side. .3 
seconds destroys most of the IO clustering Linux will do to IO, hurting IO 
performance significantly. .3 seconds keeps the disks and SATA from using 
power save mode, costing more power. (And "more power use" is not just an 
issue for machines on battery, it's *very* relevant to servers and even 
desktop machines)

Comment 3 Phil Knirsch 2007-08-13 12:04:17 EDT
That sounds really very wrong.

Peter Jones is ripping out the code upstream now, will be building new packages
once a new version is out.

Read ya, Phil
Comment 4 Till Maas 2007-09-13 01:21:38 EDT
A new package for devel is available, that should fix this bug.
Comment 5 Till Maas 2008-05-18 05:21:31 EDT
I assume this is fixed, please reopen in case it is not.