Hello,
I've been using UML for production servers for years, but I'm now
experiencing some performance problems with UBD (I've always used it
with CONFIG_BLK_DEV_UBD_SYNC).
I tried to patch UBD to run several I/O threads that perform the
requests in parallel. It improved performances significantl
Hi,
Jeff Dike wrote:
>
> Try doubling or quadrupuling the host's /proc/sys/vm/max_map_count.
Thanks a lot Jeff, it perfectly did the trick.
Just curious, is it a global or a per-process limit?
Thanks,
Nicolas
-
This SF.
stand this message? Is there a way I
can fix the problem?
Regards,
Nicolas Boullis
-
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin
Paulo da Silva wrote:
> Hi!
>
> I am using UML for a long time. It always took some time to boot.
>
> But now, after moving to 2.6.25 kernel it takes no less than 15 minutes
> to boot!!!
> In this time, my laptop (4Gb RAM Intel core duo 2.4GHz) does not make
> any relevant use of the CPU and th
would fix
the problem, as it would avoid read caches. Do you think it would be
worthy to add an option to the ubd setup to open the file/device with
O_DIRECT?
Cheers,
Nicolas Boullis
-
This SF.net email is sponsored by: Mi
Hi,
Is it possible tu use software suspend in a UML guest? I'd like to use
it to migrate a running UML guest from a host to a different host. Or
does any one know another way to migrate a running UML guest?
Cheers,
Nicolas Bo
Hi,
Jeff Dike wrote:
> On Wed, Apr 04, 2007 at 06:13:45PM +0200, Nicolas Boullis wrote:
>
>>OK, I just built a 2.6.20.4 UML kernel, and at first sight it seems to
>>run fine with size=2048M. (With higher values, it still seems fine with
>>size=2800M, and completely fails
Jeff Dike wrote:
> On Wed, Apr 04, 2007 at 11:58:06AM +0200, Nicolas Boullis wrote:
>
>>Linux version 2.6.18 ([EMAIL PROTECTED]) (gcc version 4.1.2 20061028
>
>
> Can you try something a bit more modern? I don't know that I
> specifically fixed anything related to
Jeff Dike wrote:
>>It seems that UML does not run properly when given too much memory.
>>After a few tries, there seem to be a limit somewhere below 1.5GB
>>memory. The actual limit appears to be moving with some parameters
>>(disks, root, initrd, ...).
>>With my few tests:
>> size=1408M seems to
fails
Is it something known? Is there a known maximum safe value?
Nicolas Boullis
PS: using debian etch on both the host and the guest, with
linux-image-2.6.18-4-686-bigmem 2.6.18.dfsg.1-12 on the host, and
user-mode-linux 2.6.18-1um-2
Hi,
I'd like to run some UML guests with their clocks synchronized to
external references. Of course, I thought about NTP...
My question is: is it enough to run NTP on the host and expect the
guests to keep synchronized? Or should I run NTP on each guest?
And more generally, how are clocks (both
Blaisorblade wrote:
>
> You require the host admin to assign new IPs to the host, so it cannot be
> directly supported (I mean, even if we merge the code the end user must still
> do a lot of manual setup).
That's right.
I knew I would have to set up the IP beforehand, and that would not be a
p
Hi,
The port channel looks like a good way to give remote access to the
consoles of my UML guests.
Something like "con0=fd:0,fd:1 con=port:9000" would be almost fine.
My goal is to have a pool of hosts that run a (larger) pool of guests.
Then, if I want to connect to the console of a specific gue
Hi,
I'm trying to use the standard console= kernel option to redirect the
kernel logs to 2 different consoles. Hence, I added "console=tty0
console=tty1 to my command-line.
Unfortunately, the UML kernel complains:
line_setup failed to parse "sole=tty0"
line_setup failed to parse "sole=tty1"
I
Jason Lunz wrote:
> On Tue, Nov 14, 2006 at 11:37:15AM -0500, Jeff Dike wrote:
>
>>This address range is documented - look for assignment of MAC ranges
>>to NIC manufacturers. This range will be documented as reserved for
>>private use or something similar.
>
>
> Half of all MAC-48 addresses ar
Hi,
I use UML with the tuntap virtual network driver. By default it uses
ethernet fe:fd:xx:xx:xx:xx addresses. I asked google about those
addresses with no success.
Are these addresses allocated to anyone? Reserved for experimental use?
Reserved for private use? Is it documented somewhere?
Shoul
Hi,
I've been using a few UML hosts based on a 2.6.15.2 kernel for a few
month. I was using them on a old 2.4.29-based host.
I had some segfaults that disappeared while I killed /lib/tls.
I was told on this list that these problems were because my host was
running a 2.4 kernel.
Recently, I got
n the configuration?
If it is supposed to work, would it be possible to add the otion to the
configuration for 2.6.17 (or 2.6.18 if it's too late)? I think iSCSI is
as useful as nbd, which seems to be supported.
Thanks,
Nicolas Boullis
PS: I'm not subscribed to this list; please CC
Jeff Dike wrote:
> On Wed, Mar 15, 2006 at 02:54:49PM +0100, Nicolas Boullis wrote:
>
>>I'm trying to use a vanilla 2.6.15.4 uml kernel with a debian sarge
>>installation. Most things work fine, but I get a few segfaults with a
>>few programs, such as host, nslooku
Hi,
I'm trying to use a vanilla 2.6.15.4 uml kernel with a debian sarge
installation. Most things work fine, but I get a few segfaults with a
few programs, such as host, nslookup or slapd...
For example:
# host user-mode-linux.sourceforge.net
Segmentation fault
# strace host user-mode-linux.sour
20 matches
Mail list logo
