Re: [uml-user] UML randomly hanging

2015-08-25 Thread Jon Foster
I tried the latest LTS kernel (3.18.20) and I don't see any appreciable
difference. Although maybe not the hard freeze I was seeing, it still
doesn't seem to be able to handle a multi-tasking load like running
Drupal on Apache in the UML. Hit it with 40 simultaneous connections for
a minute and it will never (at least not within a usable timespan)
recover. The symptoms are:

  1. LF (in the UML) of 75 after 15 to 20 mins of no external activity.
It never
  simmered down. Appears that the workers are not actually running.

  2. A bunch of threads running on the host with at most 2% CPU.

  3. Slow response on the pseudo console (con1=fd:0,fd:1) or SSH sessions.
  "top" and "htop" update slowly hanging for several seconds at a time.

I did some digging because the book and other sources mention "ncpus"
controls the number of simultaneous threads it will allow to run on the
host. When I configured the kernel I never found an SMP option in
menuconfig. The resultant binary didn't mention "ncpus" as a CLI
parameter. But I found the note in Kconfig that mentions it only works
in TT mode, not with SKAS, probably does more harm then good and is most
likely broken. Would be nice if the docs in the kernel tree were up to date.

In short it doesn't appear that UML can really support more than one job
at a time and I need to move onto a full virtualization solution. A
shame really I was hopping to go light on this.

THX - Jon


Richard Weinberger wrote:
> On Fri, Aug 14, 2015 at 8:25 PM, Jon Foster
>  wrote:
>   
>> I'm using a stock Debian 7 AMD-64 SMP install on a 8 core Intel Xeon
>> server, with 24GB of RAM. I'm using Debian's supplied UML kernel and
>> tools. "uname -a", from within the UML returns: "Linux lamp1 3.2.54 #2
>> Thu Feb 6 22:33:28 UTC 2014 x86_64 GNU/Linux". Using a command like:
>>
>> linux.uml "umid=$MACHNAME" con0=null,fd:2 con1=fd:0,fd:1 \
>>  con=null ssl=null "mem=4096m" "eth0=tuntap,tap0" \
>>  hostfs=/var/local/umlhostfs "ubd0=$ROOT" "ubd1=$SWAP"
>>
>> Obviosuly the environment variables are filled with appropriate values.
>>
>> I'm running a Debian 5 LAMP stack inside the UML. "/tmp" is a regular
>> disk backed filesystem. It runs ok if its not doing anything. But even
>> then from time to time you'll get a second or two pause. If it gets  a
>> modest amount of traffic it will hang for a few minutes, and things just
>> get worse and worse from there.
>>
>> While its idle I've seen "hrtimer" warnings with 9 or 10 digit
>> nanosecond counts. I think this is at the heart of my problem.
>>
>> Anybody have any advice?
>> 
>
> Can you please give a more recent kernel a try?
> 3.2. is really old.
>
>   

-- 
Sent from my Debian Linux workstation -- http://www.debian.org/intro/about

Jon Foster
JF Possibilities, Inc.
j...@jfpossibilities.com
541-410-2760
Making computers work for you!


--
___
User-mode-linux-user mailing list
User-mode-linux-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user


Re: [uml-user] UML randomly hanging

2015-08-15 Thread Richard Weinberger
On Fri, Aug 14, 2015 at 8:25 PM, Jon Foster
 wrote:
> I'm using a stock Debian 7 AMD-64 SMP install on a 8 core Intel Xeon
> server, with 24GB of RAM. I'm using Debian's supplied UML kernel and
> tools. "uname -a", from within the UML returns: "Linux lamp1 3.2.54 #2
> Thu Feb 6 22:33:28 UTC 2014 x86_64 GNU/Linux". Using a command like:
>
> linux.uml "umid=$MACHNAME" con0=null,fd:2 con1=fd:0,fd:1 \
>  con=null ssl=null "mem=4096m" "eth0=tuntap,tap0" \
>  hostfs=/var/local/umlhostfs "ubd0=$ROOT" "ubd1=$SWAP"
>
> Obviosuly the environment variables are filled with appropriate values.
>
> I'm running a Debian 5 LAMP stack inside the UML. "/tmp" is a regular
> disk backed filesystem. It runs ok if its not doing anything. But even
> then from time to time you'll get a second or two pause. If it gets  a
> modest amount of traffic it will hang for a few minutes, and things just
> get worse and worse from there.
>
> While its idle I've seen "hrtimer" warnings with 9 or 10 digit
> nanosecond counts. I think this is at the heart of my problem.
>
> Anybody have any advice?

Can you please give a more recent kernel a try?
3.2. is really old.

-- 
Thanks,
//richard

--
___
User-mode-linux-user mailing list
User-mode-linux-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user


[uml-user] UML randomly hanging

2015-08-14 Thread Jon Foster
I'm using a stock Debian 7 AMD-64 SMP install on a 8 core Intel Xeon
server, with 24GB of RAM. I'm using Debian's supplied UML kernel and
tools. "uname -a", from within the UML returns: "Linux lamp1 3.2.54 #2
Thu Feb 6 22:33:28 UTC 2014 x86_64 GNU/Linux". Using a command like:

linux.uml "umid=$MACHNAME" con0=null,fd:2 con1=fd:0,fd:1 \
 con=null ssl=null "mem=4096m" "eth0=tuntap,tap0" \
 hostfs=/var/local/umlhostfs "ubd0=$ROOT" "ubd1=$SWAP"

Obviosuly the environment variables are filled with appropriate values.

I'm running a Debian 5 LAMP stack inside the UML. "/tmp" is a regular
disk backed filesystem. It runs ok if its not doing anything. But even
then from time to time you'll get a second or two pause. If it gets  a
modest amount of traffic it will hang for a few minutes, and things just
get worse and worse from there.

While its idle I've seen "hrtimer" warnings with 9 or 10 digit
nanosecond counts. I think this is at the heart of my problem.

Anybody have any advice?

TIA - Jon

-- 
Sent from my Debian Linux workstation -- http://www.debian.org/intro/about

Jon Foster
JF Possibilities, Inc.
j...@jfpossibilities.com
541-410-2760
Making computers work for you!


--
___
User-mode-linux-user mailing list
User-mode-linux-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user