Xen users - Sep 2006 - what do you recommend for cluster fs ??

I was wondering what people use / recommend for a cluster fs for use with
xen ?

Also with this cluster fs, will the domU''s be using flat files / lvm ?
(it
should be the former as per my guess since these flat files will be stored
on the cluster fs, but then won''t performance hit will be there for
these
domU''s ? )

I don''t want to invest in a san/ nas solution, just looking for cost
effective. Right now i have a single server which i will use for the cluster
fs and slowly add more servers to the cluster as and when i need more space.
Atleast thats what i have thought till now.

All ideas are welcome.

-- 
regards,

Anand Gupta

Anand Gupta wrote:
> I was wondering what people use / recommend for a cluster fs for use 
> with xen ?
I''m currently using a firewire external hard disk. When used with the 
correct IDE-FW-chipset (Oxford are the only ones working), 4 hosts can 
connect to one disk concurrently.

The other cheap possibility is to use iSCSI (e.g. iet). Or NFS.

Harald

Anand Gupta wrote:
> I was wondering what people use / recommend for a cluster fs for use 
> with xen ?
> 
> Also with this cluster fs, will the domU''s be using flat files /
lvm ?
> (it should be the former as per my guess since these flat files will be 
> stored on the cluster fs, but then won''t performance hit will be
there
> for these domU''s ? )
You could also use CLVM to have your guests living on logical
volumes that are visible on all cluster hosts.

-- 
"You don''t have to be crazy to do this ... but it helps." --
Bob Ross

On Sep 23, 2006, at 7:44 AM, Anand Gupta wrote:
> I was wondering what people use / recommend for a cluster fs for  
> use with xen ?
We use GFS.
> Also with this cluster fs, will the domU''s be using flat files /  
> lvm ? (it should be the former as per my guess since these flat  
> files will be stored on the cluster fs, but then won''t performance
> hit will be there for these domU''s ? )
We use CLVM.
> I don''t want to invest in a san/ nas solution, just looking for  
> cost effective.
You can actually get *both* a SAN/cost effective solution from these  
folks:

   www.coraid.com

We''re very impressed with what they''re offering.

Works as advertised!
> Right now i have a single server which i will use for the cluster  
> fs and slowly add more servers to the cluster as and when i need  
> more space. Atleast thats what i have thought till now.
You need to be careful with clustering. We''re using it for high  
availability and scalability, but if you make the wrong decisions,  
you''ll end up with lower reliability.

-- 
-- Tom Mornini


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

Harald Kubota wrote:
> Anand Gupta wrote:
>> I was wondering what people use / recommend for a cluster fs for use
>> with xen ?
> I''m currently using a firewire external hard disk. When used with
the
> correct IDE-FW-chipset (Oxford are the only ones working), 4 hosts can
> connect to one disk concurrently.
> 
> The other cheap possibility is to use iSCSI (e.g. iet). Or NFS.
> 
> Harald
Doesn''t really answer the "cluster FS" question, since
that''s just
shared storage. :)

GNBD + CLVM + GFS here.

-- 
Christopher G. Stach II

Christopher G. Stach II wrote:
> Harald Kubota wrote:
>   
>> Anand Gupta wrote:
>>     
>>> I was wondering what people use / recommend for a cluster fs for
use
>>> with xen ?
>>>       
>> I''m currently using a firewire external hard disk. When used
with the
>> correct IDE-FW-chipset (Oxford are the only ones working), 4 hosts can
>> connect to one disk concurrently.
>>
>> The other cheap possibility is to use iSCSI (e.g. iet). Or NFS.
>>     
> Doesn''t really answer the "cluster FS" question, since
that''s just
> shared storage. :)
>
>   
Good point and of course true. I skipped the "cluster fs" part
completely.

Now if the question were "What do you use for shared storage to be able 
to use
Xen migration", then I''d be ok.

Harald

Hi,

anybody using or testing ocfs2?

regards Martin


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

I''m about to, actually. I''m going to make some rather easy to
deploy
cluster-in-a-box image collections and some simple scripts to setup
networking to get them up and running.

I''ll be using Pound [ http://www.apsis.ch/pound/ ] over a farm skinny
centos4 / lighttpd / fastcgi dom-u''s on top of a conventional MySQL
cluster for db, with failover pound load balancer and mysql manager
dom-u''s (located on different machines), just based on the common 3.0.x
para-virtualized kernel. 

Its use will be easily rolling out something to handle high traffic or
image / media intense web sites, forums that take off quickly or sites
prone to frequent /.''ings or diggs.

I''ve yet to get ocfs2 working but its mostly due to lack of effort on
my
part, I just don''t have the time to work on hobby stuff anymore.

If you wouldn''t mind posting your experiences with it should you elect
to try it, I''d appreciate the reading :) I often just default to gfs
because it works and I can deploy it quickly.

It seems like a cinch .. check : [ from oracle''s site
http://oss.oracle.com/projects/ocfs2/ ] :

>> Ubuntu: OCFS2 is included in the stock Dapper kernel. Ubuntu users 
>> must also install the ocfs2-tools and ocfs2console packages, which
>> are in Dapper as well. Thanks goes to Fabio Massimo Di Nitto for 
>> doing the packaging work.
Best,
-Tim

On Sun, 2006-09-24 at 19:03 +0200, Martin Hierling
wrote:
> Hi,
> 
> anybody using or testing ocfs2? 
> 
> regards Martin
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users

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

Tom Mornini wrote:
> On Sep 23, 2006, at 7:44 AM, Anand Gupta wrote:
>
>> I was wondering what people use / recommend for a cluster fs for use 
>> with xen ?
>
> We use GFS.
>
>> Also with this cluster fs, will the domU''s be using flat files
/ lvm
>> ? (it should be the former as per my guess since these flat files 
>> will be stored on the cluster fs, but then won''t performance
hit will
>> be there for these domU''s ? )
>
> We use CLVM.
I hear talk of CLVM, but I''m trying to understand what the difference
is
between CLVM and normal LVM, where the CLVM software is and what I have 
to do to turn my LVM system into a CLVM?????

I''d also be interested to know if you can combine CLVM with a dm-raid 
solution, to create a clustered raid device? I understand that 
dm-raid0/1 has been released, and there''s a dm-raid4/5 in alpha at the 
moment.

Thanks,

Tony Wright.

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

On Sep 24, 2006, at 11:39 AM, Anthony Wright wrote:
> Tom Mornini wrote:
>
>> On Sep 23, 2006, at 7:44 AM, Anand Gupta wrote:
>>
>>> I was wondering what people use / recommend for a cluster fs for  
>>> use with xen ?
>>
>> We use GFS.
>>
>>> Also with this cluster fs, will the domU''s be using flat
files /
>>> lvm ? (it should be the former as per my guess since these flat  
>>> files will be stored on the cluster fs, but then won''t  
>>> performance hit will be there for these domU''s ? )
>>
>> We use CLVM.
>
> I hear talk of CLVM, but I''m trying to understand what the  
> difference is between CLVM and normal LVM, where the CLVM software  
> is and what I have to do to turn my LVM system into a CLVM?????
It''s part of the Red Hat Cluster Suite.
> I''d also be interested to know if you can combine CLVM with a dm- 
> raid solution, to create a clustered raid device? I understand that  
> dm-raid0/1 has been released, and there''s a dm-raid4/5 in alpha at
> the moment.
I''m quite certain that dm-raid doesn''t work in a clustered  
environment yet.

I have a feeling you''re anxious (as are many) for a solution like this:

http://sourceware.org/cluster/ddraid/

Unfortunately there''s no date on the article, and it hasn''t
been
updated in months.

If you''re looking for something useful *today*, then I''d
recommend a
SAN solution. The AoE SAN solution from Coraid that I mentioned  
earlier isn''t much more expensive than plain ATA disks, so
it''s not
prohibitive.

-- 
-- Tom Mornini


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

Martin Hierling wrote:
> Hi,
> 
> anybody using or testing ocfs2?
I''ve used ocfs2 a little under Gentoo. I was using evms + heartbeat to
manage block devices shared via AoE. It all seems to work ok, but I have
to seem to have to restart heartbeat occasionally because the evms
client want connect to the cluster manager properly.

Cheers,

Brad

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

Do you happen to know if ocfs2 dynamically allocates inodes similar to
how GFS does things? Or is it more ext3-ish in behavior?

With gfs, its not uncommon to see 100% of your inodes in use.. where
with ext3 one would wonder if the world was about to end abruptly.

This can really throw network / heartbeat monitors for a loop if you
aren''t expecting it.

I found that out with gfs, and figured I''d save all who may try ocfs2
the 8 or 9 hours it took to figure out just why it was happening. I was
looking for bug reports on xensource when It was Red Hat''s tree that I
should have been climbing. :)

HTH
-Tim

On Mon, 2006-09-25 at 15:07 +1000, Brad Plant wrote:
> Martin Hierling wrote:
> > Hi,
> > 
> > anybody using or testing ocfs2?
> 
> I''ve used ocfs2 a little under Gentoo. I was using evms +
heartbeat to
> manage block devices shared via AoE. It all seems to work ok, but I have
> to seem to have to restart heartbeat occasionally because the evms
> client want connect to the cluster manager properly.
> 
> Cheers,
> 
> Brad
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
> 
> 

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

Tim Post wrote:
> Do you happen to know if ocfs2 dynamically allocates inodes similar to
> how GFS does things? Or is it more ext3-ish in behavior?
To be honest, I have no idea. But I did find this with a quick google:
http://oss.oracle.com/projects/ocfs2/dist/documentation/fasheh.pdf
<quote>
File metadata is allocated in blocks via a sub
allocation mechanism. All block allocators in
OCFS2 grow dynamically. Most notably, this
allows OCFS2 to grow inode allocation on demand.
</quote>
> With gfs, its not uncommon to see 100% of your inodes in use.. where
> with ext3 one would wonder if the world was about to end abruptly.
> 
> This can really throw network / heartbeat monitors for a loop if you
> aren''t expecting it.
> 
> I found that out with gfs, and figured I''d save all who may try
ocfs2
> the 8 or 9 hours it took to figure out just why it was happening. I was
> looking for bug reports on xensource when It was Red Hat''s tree
that I
> should have been climbing. :)
Thanks for the heads up :-)

Cheers,

Brad

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

Tim,
setup is easy (about 1h) but my test szenario was: P3-1GHz with Dom0 as
iSCSI Target and 3 DomU as ocfs2 "clients" mounting the same FS over
open-iscsi. That works out of the box but under heavy load the Clients get
kicked by the heardbeat. That obviously was because the complete network was
fuff of iscsi traffic (bonnie). I will test that in a real environment in
about 2 weeks. will reoport here.

regards Martin


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

On Sep 25, 2006, at 8:09 AM, Martin Hierling wrote:
> setup is easy (about 1h) but my test szenario was: P3-1GHz with  
> Dom0 as iSCSI Target and 3 DomU as ocfs2 "clients" mounting the  
> same FS over open-iscsi. That works out of the box but under heavy  
> load the Clients get kicked by the heardbeat. That obviously was  
> because the complete network was fuff of iscsi traffic (bonnie). I  
> will test that in a real environment in about 2 weeks. will reoport  
> here.
You might get better performance and leave this problem behind with  
AoE as it''s much less demanding under very high loads.

-- 
-- Tom Mornini


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

On Mon, Sep 25, 2006 at 09:52:19AM -0700, Tom Mornini
wrote:
> On Sep 25, 2006, at 8:09 AM, Martin Hierling wrote:
> 
> >setup is easy (about 1h) but my test szenario was: P3-1GHz with  
> >Dom0 as iSCSI Target and 3 DomU as ocfs2 "clients" mounting
the
> >same FS over open-iscsi. That works out of the box but under heavy  
> >load the Clients get kicked by the heardbeat. That obviously was  
> >because the complete network was fuff of iscsi traffic (bonnie). I  
> >will test that in a real environment in about 2 weeks. will reoport  
> >here.
> You might get better performance and leave this problem behind with  
> AoE as it''s much less demanding under very high loads.
I must admit that it sounds like you''re running your frontend and
backend networks on the same LAN. 

It''s probably wise to use a dedicated network for storage network
traffic, because as you''ve seen, it can quite easily saturate the link,
causing performance issues. 

It''s also entirely possible that moving the storage onto a dedicated
network would bring security benefits, assuming you''ve taken proper
precautions (there''s no need for it to have access / be accessible from
the internet). 
-- 
Ceri Storey <cez@necrofish.org.uk>
''What I really want is "apt-get smite"'' 
    --Rob Partington <http://rjp.frottage.org>

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

On Sep 25, 2006, at 10:00 AM, Ceri Storey wrote:
> On Mon, Sep 25, 2006 at 09:52:19AM -0700, Tom Mornini wrote:
>> On Sep 25, 2006, at 8:09 AM, Martin Hierling wrote:
>>
>>> setup is easy (about 1h) but my test szenario was: P3-1GHz with
>>> Dom0 as iSCSI Target and 3 DomU as ocfs2 "clients"
mounting the
>>> same FS over open-iscsi. That works out of the box but under heavy
>>> load the Clients get kicked by the heardbeat. That obviously was
>>> because the complete network was fuff of iscsi traffic (bonnie). I
>>> will test that in a real environment in about 2 weeks. will reoport
>>> here.
>
>> You might get better performance and leave this problem behind with
>> AoE as it''s much less demanding under very high loads.
>
> I must admit that it sounds like you''re running your frontend and
> backend networks on the same LAN.
>
> It''s probably wise to use a dedicated network for storage network
> traffic, because as you''ve seen, it can quite easily saturate the
> link,
> causing performance issues.
>
> It''s also entirely possible that moving the storage onto a
dedicated
> network would bring security benefits, assuming you''ve taken
proper
> precautions (there''s no need for it to have access / be accessible
> from
> the internet).
These are all excellent points. We have separate IP and AoE networks.

-- 
-- Tom Mornini


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

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

What advantages are there to using a FS via iSCSI instead of just using
NFS? Are there some websites/articles you can point me to?

- --tod
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFFGBnX6xsamcQR838RAg+wAKDaYq4wqXU7DIbheSxvo2Ak+IbvXwCgo3Zt
LrG7s3xtaZMQ7++MBOo+BNc=KVY+
-----END PGP SIGNATURE-----

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

If you are going through the trouble of a second network, wouldn''t
using
1Gbit fibre make more sense?  1gbt fibre-channel is actually cheaper than 
gigabit Ethernet, assuming you are buying name-brand equipment, 
comparing used to used,  and for disk use, fibre-channel is much faster 
than going over the network.  It also has basic disk 
organization/pesudo-security built in already.

I use pci-x qlogic 2200 cards (around $10 per)  and Brocade silworm 2800 
switches (around $100/each)  along with whatever fibre arrays I can find 
(I have a IBM EXP-500 right now-  Nice!  but it was $250. 
You can get used dell/EMC 10 bay half-height arrays for little more than 
shipping; but that''s ''cause they are flimsy crap.  After
getting it
shipped with drives in it, you will have bad/flaky slots.    I just 
ordered a Sun StorEdge A5200 for around $150, but those are low-profile, 
and the half-height drives are extremely cheap-  you can get 10KRPM 
half-height 73G drives for around $10/each.  that goes up to $50 or so 
for the low-profile drives of the same spec, and you only get a 30% 
density improvement.)


On Mon, 25 Sep 2006, Tom Mornini wrote:
> Date: Mon, 25 Sep 2006 10:59:39 -0700
> From: Tom Mornini <tmornini@engineyard.com>
> To: Xen Users <Xen-users@lists.xensource.com>
> Cc: Martin Hierling <martin@mh-itc.de>
> Subject: Re: [Xen-users] what do you recommend for cluster fs ??
> 
> On Sep 25, 2006, at 10:00 AM, Ceri Storey wrote:
>
>> On Mon, Sep 25, 2006 at 09:52:19AM -0700, Tom Mornini wrote:
>>> On Sep 25, 2006, at 8:09 AM, Martin Hierling wrote:
>>> 
>>>> setup is easy (about 1h) but my test szenario was: P3-1GHz with
>>>> Dom0 as iSCSI Target and 3 DomU as ocfs2 "clients"
mounting the
>>>> same FS over open-iscsi. That works out of the box but under
heavy
>>>> load the Clients get kicked by the heardbeat. That obviously
was
>>>> because the complete network was fuff of iscsi traffic
(bonnie). I
>>>> will test that in a real environment in about 2 weeks. will
reoport
>>>> here.
>> 
>>> You might get better performance and leave this problem behind with
>>> AoE as it''s much less demanding under very high loads.
>> 
>> I must admit that it sounds like you''re running your frontend
and
>> backend networks on the same LAN.
>> 
>> It''s probably wise to use a dedicated network for storage
network
>> traffic, because as you''ve seen, it can quite easily saturate
the link,
>> causing performance issues.
>> 
>> It''s also entirely possible that moving the storage onto a
dedicated
>> network would bring security benefits, assuming you''ve taken
proper
>> precautions (there''s no need for it to have access / be
accessible from
>> the internet).
>
> These are all excellent points. We have separate IP and AoE networks.
>
> -- 
> -- Tom Mornini
>
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Well, if ATA meets your needs, that''s fine;  last time I used 7.5K
SATA, I
had long "pauses" whenever all the other computers on a particular 
disk/array ran their daily crontab, and other times of even moderately 
high IO.  On fibre, everything is pretty smooth.  under load it can get 
slow, I''m running on 2.0Ghz Xeons w/ a 400Mhz bus and pc2100 ram, but
the
latencies are always low;  the system is always responsive.

Also, all the external SATA chassis have been anything but commodity, 
last time I looked (which admittedly was a while ago)  both 3-ware and 
Adaptec had competing, incompatible "standards" for their 4-lane 
connectors, and as I recall, at the time the RAID cards almost cost as 
much as the disks in the 14 bay supermicro we used.

Are there standard interconnects now, such that I can buy a disk chassis 
from one company and be fairly certain it will connect to a storage 
controller from another?  (I really would like to know...  I would like to 
try out some of those 2.5" 10KRPM  SAS drives-  those look cool.  but I 
refuse to do so until there is an open standard supported by more than 
one vendor.)  I have seen fairly standard-looking sata JBOD cases that 
used fibre channel interconnects;  I will probably be buying some for 
storage shortly;


Personally, I place zero value on "support" as even from the premium 
vendors like EMC, and even when you escalate up to Engineering, you
don''t
get anyone that knows more than I do, and it takes days to get to that 
point.

However, good manufacturer warranties are really nice-  I''ll pay a
significant premium for those.  I always buy corsair ram for that reason; 
awesome warranty.  and new disks are nearly always from Seagate; Seagate 
has excellent warranty support.  You don''t even have to talk to a
person;
just fill out a webform (easy when you have a barcode scanner)  and mail 
them in.


So yeah, if there are standard SAS/SATA interconnects, I''d like to know
about them, as that seems to be where Seagate thinks things are going. 
(and I like the way SAS scales.  one bus per disk when you have as 
many spindles as I do would equal some really nice thruput..  that''s 
the only thing the fibre disks lack.  They are okay under heavy random 
load where SATA chokes, but they are also only okay under sequential load, 
where SATA flies.  Of course, I don''t see much sequential activity 
(sequential activity from several hosts to the same disk equals random 
load on the disk) so I go with the fibre.  Still, good sequential 
performance would make full system restores and a few other things run a 
whole lot faster.  you don''t need to do a full system restore very
often,
but when you do, you really, really need to do it. )
> On Sep 25, 2006, at 11:17 AM, Luke Crawford wrote:
>
>> If you are going through the trouble of a second network,
wouldn''t using
>> 1Gbit fibre make more sense?  1gbt fibre-channel is actually cheaper
than
>> gigabit Ethernet, assuming you are buying name-brand equipment,
comparing
>> used to used,  and for disk use, fibre-channel is much faster than
going
>> over the network.  It also has basic disk organization/pesudo-security 
>> built in already.
>
> Not for me. The way I see things, with Coraid and AoE, I get to stay in 
> uber-commodity land with SATA disks, GbE cards and switches. The current
AoE
> drivers balance requests over multiple ports, so in a fully redundant 
> configuration you get nearly 2Gbps throughput.
>
>> I use pci-x qlogic 2200 cards (around $10 per)  and Brocade silworm
2800
>> switches (around $100/each)  along with whatever fibre arrays I can
find (I
>> have a IBM EXP-500 right now-  Nice!  but it was $250. You can get used
>> dell/EMC 10 bay half-height arrays for little more than shipping; but 
>> that''s ''cause they are flimsy crap.  After getting it
shipped with drives
>> in it, you will have bad/flaky slots.    I just ordered a Sun StorEdge 
>> A5200 for around $150, but those are low-profile, and the half-height 
>> drives are extremely cheap-  you can get 10KRPM half-height 73G drives
for
>> around $10/each.  that goes up to $50 or so for the low-profile drives
of
>> the same spec, and you only get a 30% density improvement.)
>
> I like the fact that the stuff I''m buying is reasonably
inexpensive, and is
> brand new with manufacturer warranties.
>
> Additionally, I like the fact that it''s all headed in the right
direction for
> 10GE soon, when those prices drop as well.
>
> -- 
> -- Tom Mornini
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sep 25, 2006, at 1:46 PM, Luke Crawford wrote:
> Well, if ATA meets your needs, that''s fine;  last time I used 7.5K
> SATA, I had long "pauses" whenever all the other computers on a  
> particular disk/array ran their daily crontab, and other times of  
> even moderately high IO.  On fibre, everything is pretty smooth.   
> under load it can get slow, I''m running on 2.0Ghz Xeons w/ a
400Mhz
> bus and pc2100 ram, but the latencies are always low;  the system  
> is always responsive.
http://www.coraid.com

So far, so good, and I''m not the only one using them.

-- 
-- Tom Mornini



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

On Mon, 25 Sep 2006, Tom Mornini wrote:
> http://www.coraid.com
Their SR1520 looks almost exactly like the SuperMicro I was describing 
having used before.  Interesting stuff.   do you use the AOE in dom0 and 
export block devices?  or do you do AoE directly from within the DomUs?

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

On Sep 25, 2006, at 3:31 PM, Luke Crawford wrote:
> On Mon, 25 Sep 2006, Tom Mornini wrote:
>> http://www.coraid.com
>
> Their SR1520 looks almost exactly like the SuperMicro I was  
> describing having used before.  Interesting stuff.   do you use the  
> AOE in dom0 and export block devices?  or do you do AoE directly  
> from within the DomUs?
It *is* a SuperMicro, but with their software installed. :-)

We use AoE to create the block devices in Dom0, then export them with  
the DomUs, as you guessed.

If you want to work with AoE, make sure you get their drivers.  
They''re in-kernel now, but for whatever reason, they''re very
far out
of date in the standard kernel sources.

-- 
-- Tom Mornini


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

Could you please direct me some documentation/howto where i can get a feel
of it and see how to go about installing and using it ? I heard about the RH
Cluster Suite, however wasn''t able to find any documentation to use it.

On 9/23/06, Rik van Riel <riel@redhat.com> wrote:
>
> Anand Gupta wrote:
> > I was wondering what people use / recommend for a cluster fs for use
> > with xen ?
> >
> > Also with this cluster fs, will the domU''s be using flat
files / lvm ?
> > (it should be the former as per my guess since these flat files will
be
> > stored on the cluster fs, but then won''t performance hit will
be there
> > for these domU''s ? )
>
> You could also use CLVM to have your guests living on logical
> volumes that are visible on all cluster hosts.
>
> --
> "You don''t have to be crazy to do this ... but it
helps." -- Bob Ross
>


-- 
regards,

Anand Gupta

On Thu, Sep 28, 2006 at 01:23:21AM +0530, Anand Gupta
wrote:
> Could you please direct me some documentation/howto where i can get a feel
> of it and see how to go about installing and using it ? I heard about the
RH
> Cluster Suite, however wasn''t able to find any documentation to
use it.
There''s some docs linked off the Clustersuite dev pages:

http://sources.redhat.com/cluster/

Regards,
Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

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

Thanks for the link. Already tried looking there, however was confused. I
guess there is no other way then to try to find way in those docs itself.

On 9/28/06, Daniel P. Berrange <berrange@redhat.com>
wrote:
>
> On Thu, Sep 28, 2006 at 01:23:21AM +0530, Anand Gupta wrote:
> > Could you please direct me some documentation/howto where i can get a
> feel
> > of it and see how to go about installing and using it ? I heard about
> the RH
> > Cluster Suite, however wasn''t able to find any documentation
to use it.
>
> There''s some docs linked off the Clustersuite dev pages:
>
> http://sources.redhat.com/cluster/
>
> Regards,
> Dan.
> --
> |=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496
> -=|
> |=-           Perl modules: http://search.cpan.org/~danberr/
>               -=|
> |=-               Projects: http://freshmeat.net/~danielpb/
> -=|
> |=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B
> 9505  -=|
>


-- 
regards,

Anand Gupta

All the documentation is available on line.

http://www.redhat.com/docs/manuals/csgfs/


On Thu, 2006-09-28 at 01:23 +0530, Anand Gupta wrote:
> Could you please direct me some documentation/howto where i can get a
> feel of it and see how to go about installing and using it ? I heard
> about the RH Cluster Suite, however wasn''t able to find any
> documentation to use it. 
> 
> 
> On 9/23/06, Rik van Riel <riel@redhat.com> wrote:
> 
>         Anand Gupta wrote:
>         > I was wondering what people use / recommend for a cluster fs
>         for use
>         > with xen ?
>         >
>         > Also with this cluster fs, will the domU''s be using
flat
>         files / lvm ?
>         > (it should be the former as per my guess since these flat
>         files will be 
>         > stored on the cluster fs, but then won''t performance
hit
>         will be there
>         > for these domU''s ? )
>         
>         You could also use CLVM to have your guests living on logical
>         volumes that are visible on all cluster hosts. 
>         
>         --
>         "You don''t have to be crazy to do this ... but it
helps." --
>         Bob Ross
> 
> 
> 
> 
> -- 
> regards,
> 
> Anand Gupta 
> 
> --
> Fedora-xen mailing list
> Fedora-xen@redhat.com
> https://www.redhat.com/mailman/listinfo/fedora-xen

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

On 9/28/06, Andrew Cathrow <acathrow@redhat.com>
wrote:
>
>  All the documentation is available on line.
>
> http://www.redhat.com/docs/manuals/csgfs/
>
Thanks for the pointer Andrew.

-- 
regards,

Anand Gupta