[lkml]   [2011]   [Feb]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH v6 0/4] fadvise(DONTNEED) support
On Tue, Feb 22, 2011 at 11:03:51AM -0600, Jeffrey Hundstad wrote:
> On 02/22/2011 07:28 AM, Andrea Arcangeli wrote:
> > Hi Minchan,
> > That's my point, we should check if the "thrashing horribly" is really
> > a "recently" or if it has always happened before with 2.6.18 and previous.
> I would bet that "thrashing" isn't what he meant. He almost certainly
> meant "needless I/O" and nothing to do with swapping. I've seen this
> similar report before. This report is almost always gets answered with
> a "set your swappiness to an appropriate value" type answer. Or a "your

Hmm but swappiness can't help if "needless I/O in the morning after
backup" is the problem. So it wouldn't be the right suggestion if
"needless I/O is the problem".

We need a "vmstat 1" during the "trashing horribly"/"needless I/O" in
the morning to be sure.

> userspace app needs to get smarter" type answer. I think they are
> trying to do the latter, but need some help from the kernel, and that's
> what they're trying to do here.

That's fine thing. But I got reports indipendent of this thread from
friends, asking me how to prevent swapping overnight because of
backup, so I connected that report to the above "trashing
horribly"/recently. I had no problems personally but may backup runs
once (not twice to activate/reference pages). I had no time to
investigate this further yet to see if things works ok for me even
with a rsync loop activating and referencing everything.

 \ /
  Last update: 2011-02-22 18:17    [W:0.070 / U:3.004 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site