[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] Re: Xen 3.2.0 -> Newer kernel and more..



Hi,

Thanks for reply.


On Thu, 24 Jan 2008, Tom Brown wrote:



AFAIK, any network filesytem is going to be "slow" as compared to a
filesystem on an exclusively used virtual block device. Mount root and the rest of the
operating system on some sort of dedicated block device and only used a
shared network filesytem for the parts of the system that _need_ to be accessed from
multiple servers.

Of course. I should explain little better.

- Both Dom0's are Intel Core 2.
- Xen build from sources (3.2.0)
- LVM2.

DomU's uses ext3 for root (/) partition and OCFS2 for /www
and then of course swap partition.

Inside /www is Apache and PHP compiled from sources.
Of course some changes to apache for .pid and so on...

I have tested DRBD & OCFS2 two ways.

- DRBD & OCFS2 in Dom0 -> /dev/drbd0 shared to DomU as normal vbd.
- DRBD % OCFS2 in DomU.

!!! HOX !!!
"tests" i mention below is only "tests".
Everyone know's that those are not best way to make test. I make better test tomorrow when go work.

1)
Writing speed is actually good if writing zero to one file.
Like 'dd if=/dev/zero of=./foobar.bin bs=4096 count=40000'
-> Write speed ~60Mb/sek

But when change input to /dev/urandom speed dropped ->  ~5.7Mb/sek

2)
Downloaded apache-2.2.9.tar.gz to /www.
Extracting works good enought, but nothing compared to normal ext3 vbd.

But when delete ./httpd-2.2.9 which is around 30Mb it takes over 60sec before operation is done!!!
How can deleting be slower then "inserting" ??

Im gonna make lot's of more testing tomorrow.

If all you are trying to get is migration capabilities, then use
something like AOE or iSCSI instead.

Yea. Im aware bought of those.
It just that im trying todo "Two node load-balanced high-availability LAMP environment" Im certainly aware that i need third machine for backups and that this would be much easier with some kind of storage backend.

(Best to use an initrd so that you can actually have the domU look after
the details... which will go with it if it gets migrated)

Im sorry, but can You explain this little better?

Regards,
Kalle


On Thu, 24 Jan 2008, Tom Brown wrote:



AFAIK, any network filesytem is going to be "slow" as compared to a filesystem on an exclusively used virtual block device. Mount root and the rest of the operating system on some sort of dedicated block device and only used a shared network filesytem for the parts of the system that _need_ to be accessed from multiple servers.

If all you are trying to get is migration capabilities, then use something like AOE or iSCSI instead.

(Best to use an initrd so that you can actually have the domU look after the details... which will go with it if it gets migrated)

-Tom

On Thu, 24 Jan 2008, Kalle wrote:

Hi,

Thanks man.

I actually get that OCFS2 to work with my 2.6.18.8-xen but it was bit of a challenge.

My cluster is now up and running and need lot's of testing.
But im pretty sure that OCFS2 it's not the one im looking for, because it looks to be pretty damn slow at least on my setup.

Regards,
Kalle


On Thu, 24 Jan 2008, Christian Tramnitz wrote:

> So once again. If someone knows howto apply that 'OCFS2 add_bio patch' > to Xenified 2.6.18 kernel i would be more then happy little admin.
>  Or howto to patch 2.6.20 with Dom0 support.

 Gentoo has a 2.6.20 xenified Dom0 kernel which is running fine for me for
 a couple weeks...


 Best regards,
   Christian


 _______________________________________________
 Xen-users mailing list
 Xen-users@xxxxxxxxxxxxxxxxxxx
 http://lists.xensource.com/xen-users


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


----------------------------------------------------------------------
tbrown@xxxxxxxxxxxxx   | Courage is doing what you're afraid to do.
http://BareMetal.com/  | There can be no courage unless you're scared.
                      | - Eddie Rickenbacker


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.