Re: udevd on very large systems: again

[George Beshers <gbeshers@sgi.com>]

Kay Sievers wrote:
> On 5/2/07, Kay Sievers <kay.sievers@vrfy.org> wrote:
>> On 5/2/07, George Beshers <gbeshers@sgi.com> wrote:
>> >
>> > After some testing the following patch seems safe for dealing with 
>> computers
>> > with more than 500 processors.  It occurred to me that it might be
>> > easier and
>> > just as useful to note when udev is running on a large computer 
>> (e.g., >= 8p
>> > and >= 16Gb of memory) and simply not worry about throttling at 
>> that point.
>> >
>> > The patch should apply cleanly against udev-110.
>> >
>> > Comments welcome.
>>
>> Looks fine so far, but I still don't see how reading a seq-proc-file
>> could miss entries. Are you expecting this to happen? Did you ever see
>> this for any of these files?
>
> If I read linux/fs/proc/proc_misc.c correctly, /proc/stat fills a
> buffer of 128kB in one call and passes the data to the file.
>
> I wonder how long it will take, that you guys hit the 128k limit. :)
>
> Thanks,
> Kay
Hmmm... well on the 128p system where I did my testing the size of
/proc/stat ranged from 5317 to 5451 (that was with 151 active procs).
At that scale 2048 should be OK.

BTW.  I should have said that the log was generated with a kernel
build (make -j 128 or something like that) going.

George


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel