Mailing-List: contact cloudstack-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: cloudstack-dev@incubator.apache.org
Received-SPF: pass (athena.apache.org: domain of shadowsor@gmail.com
 designates 209.85.212.47 as permitted sender)
MIME-Version: 1.0
In-Reply-To: 
 <CAF24echgbed4eMt8Cd-iX5XoM=MhY3nA3VL15Tt6BWk=VLzmgw@mail.gmail.com>
References: 
 <CAF24ecjFakT+txCfn_oJ0+WphA5iZgZyKzVdYm+gPrBKM0pPNQ@mail.gmail.com>
	<C66C814C1ABFA8449FF65CB44953A8B4012CC09A8C0A@SJCPMAILBOX01.citrite.net>
	<CAF24ecjsLP+buV3S5HhJA-Bn2byYW-yuF_VkzJao1xJjDqOY-Q@mail.gmail.com>
	<C66C814C1ABFA8449FF65CB44953A8B4012CC09A8C0F@SJCPMAILBOX01.citrite.net>
	<97F4356AEA71904482CD192135C038F9011CC1390D89@BANPMAILBOX01.citrite.net>
	<97F4356AEA71904482CD192135C038F9011CC1390D8F@BANPMAILBOX01.citrite.net>
	<CAF24eci7WGZw-3ELKzwTva_hQY5WEJ9G-vpFPPGDio8w5qQEmA@mail.gmail.com>
	<511E3F84.9090401@widodh.nl>
	<CAF24echcYu38T5bXBHE_D731k5NzCnq2ZuvvNp2xp+qZAikixg@mail.gmail.com>
	<511E51ED.4060903@widodh.nl>
	<CAF24echgbed4eMt8Cd-iX5XoM=MhY3nA3VL15Tt6BWk=VLzmgw@mail.gmail.com>
Date: Fri, 15 Feb 2013 08:49:32 -0700
Message-ID: 
 <CALFpzo4EAoV4ozAubUU60Z+nLrwbZFqSxaraVZ+xC9hWEf3muw@mail.gmail.com>
Subject: Re: Change qcow disk cache=none
From: Marcus Sorensen <shadowsor@gmail.com>
To: cloudstack-dev@incubator.apache.org
Content-Type: multipart/alternative; boundary=f46d043d64b345ac2a04d5c5526a

--f46d043d64b345ac2a04d5c5526a
Content-Type: text/plain; charset=ISO-8859-1

Yes, lots of people will want it.  For the longest time 'none' was fastest,
and still is except for some of the newer storage like RBD.

With a little more tinkering you could apply a different setting for
block,file,and network storage as they're already split out right there in
that file.

I worry a tiny bit though that such a feature won't be exposed to end
users, and will be a way of boosting performance at the cost of
reliability/safety. Users might be aware of the performance but not realize
that their cloud provider is running writeback. That's not a reason to not
have the feature, of course.
On Feb 15, 2013 8:26 AM, "Jason Villalta" <jason@rubixnet.com> wrote:

> Thanks for all your help guys.  I am on 4.0 currently.  I will go the
> agent route and see what I can do.
>
> As for putting in the feature request.  I will do that as well since I
> am sure the more people use cloudstack the more people will run into
> this.  I really thing the UI is far superior experience.
>
> On a side note, the VLAN stuff needs some better documentation around
> configuration with KVM.  Is there a place I can add some of my notes
> once I get past this disk issue and test everything?
>
>
> On Fri, Feb 15, 2013 at 10:19 AM, Wido den Hollander <wido@widodh.nl>
> wrote:
> > On 02/15/2013 03:05 PM, Jason Villalta wrote:
> >>
> >> So is there anyway to make this work today or will I have to wait
> >> until this is properly implemented?  This may mean I need to take
> >> another look at openstack or opennebula and switch back.
> >>
> >
> > No, not today in a release of CloudStack.
> >
> > What you however can do is download the source and build the agent
> plugin.
> > Assuming you use 4.0:
> >
> > $ cd plugins/hypervisor/kvm
> > $ nano src/com/cloud/hypervisor/kvm/resource/LibvirtVMDef.java
> > $ mvn
> >
> > Change the "none" setting to what you like.
> >
> > Now you will have a .jar file in the "target" directory which you use to
> > replace the /usr/share/java/cloud-plugin-hypervisor-kvm.jar on your
> > production systems.
> >
> > Wido
> >
> >
> >> On Fri, Feb 15, 2013 at 9:00 AM, Wido den Hollander <wido@widodh.nl>
> >> wrote:
> >>>
> >>> On 02/15/2013 02:31 PM, Jason Villalta wrote:
> >>>>
> >>>>
> >>>> Is there any way easy way to make this setting dependent on the
> >>>> storage definition?  For example some primary storage may need this
> >>>> set to one value such as cache=writeback such as with the case of ZFS,
> >>>> GlusterFS, CEPH.  Then other storage may want a different setting such
> >>>> as cache=writethough or none which is the case with a local ext4
> >>>> storage or NFS.
> >>>>
> >>>
> >>> Exactly.
> >>>
> >>> I brought this up about a week ago, see this thread:
> >>> http://markmail.org/thread/qmrod55gfmhwzot4
> >>>
> >>> It shouldn't be a global property, it should be defined per storage
> pool,
> >>> or
> >>> even better, per disk offering.
> >>>
> >>> Right now we are talking about the cache setting, but tomorrow it could
> >>> be a
> >>> different setting.
> >>>
> >>> Wido
> >>>
> >>>
> >>>>
> >>>>
> >>>> On Thu, Feb 14, 2013 at 11:08 PM, Rajesh Battala
> >>>> <rajesh.battala@citrix.com> wrote:
> >>>>>
> >>>>>
> >>>>> Hi Jason Villalta,
> >>>>> In LibvirtVmDef.java, there is DiskDef class ,  get the cache value
> >>>>> from
> >>>>> the Global config, while constructing the xml in the override method
> >>>>> the
> >>>>> class toString() specify the cache value, currently it is using
> "none".
> >>>>>
> >>>>> NOTE: whenever there is change in GC value, MS should restart for the
> >>>>> value to take effect.
> >>>>>
> >>>>> Thanks
> >>>>> Rajesh Battala
> >>>>>
> >>>>>> -----Original Message-----
> >>>>>> From: Rajesh Battala [mailto:rajesh.battala@citrix.com]
> >>>>>> Sent: Friday, February 15, 2013 9:22 AM
> >>>>>> To: cloudstack-users@incubator.apache.org; cloudstack-
> >>>>>> dev@incubator.apache.org
> >>>>>> Subject: RE: Change qcow disk cache=none
> >>>>>>
> >>>>>> Hi Jason Villalta,
> >>>>>>
> >>>>>> Can you raise the ticket for the same. I can work on it if you
> assign
> >>>>>> to
> >>>>>> me.
> >>>>>>
> >>>>>> Thanks
> >>>>>> Rajesh Battala
> >>>>>>
> >>>>>>> -----Original Message-----
> >>>>>>> From: Edison Su [mailto:Edison.su@citrix.com]
> >>>>>>> Sent: Friday, February 15, 2013 4:59 AM
> >>>>>>> To: cloudstack-users@incubator.apache.org
> >>>>>>> Subject: RE: Change qcow disk cache=none
> >>>>>>>
> >>>>>>> The code is at
> https://git-wip-us.apache.org/repos/asf?p=incubator-
> >>>>>>>
> >>>>>>>
> cloudstack.git;a=blob;f=plugins/hypervisors/kvm/src/com/cloud/hypervis
> >>>>>>> or/kv
> >>>>>>>
> >>>>>>
> m/resource/LibvirtVMDef.java;h=acfd9cf1fe8eba217f09e20adf3101d1295d51d
> >>>>>>>
> >>>>>>>
> >>>>>>> a;hb=HEAD, search "cache".
> >>>>>>> you can add a global configuration in cloudstack, or add a
> >>>>>>> configuration in agent.properties, to specify cache attribute.
> >>>>>>>
> >>>>>>>> -----Original Message-----
> >>>>>>>> From: Jason Villalta [mailto:jason@rubixnet.com]
> >>>>>>>> Sent: Thursday, February 14, 2013 2:16 PM
> >>>>>>>> To: cloudstack-users@incubator.apache.org
> >>>>>>>> Subject: Re: Change qcow disk cache=none
> >>>>>>>>
> >>>>>>>> Ok, I see the folks at openstack ran into a similar issue.
> >>>>>>>> https://bugs.launchpad.net/nova/+bug/959637
> >>>>>>>>
> >>>>>>>> How can I contribute to this?  Can you recommend a part of the
> code
> >>>>>>>> to look at so I can build and test a patch?
> >>>>>>>>
> >>>>>>>> On Thu, Feb 14, 2013 at 5:13 PM, Edison Su <Edison.su@citrix.com>
> >>>>>>>> wrote:
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>> need to change cloudstack java code, right now, it's hard coded
> to
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> "cache=none".
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>
> >>>>>>>>>> -----Original Message-----
> >>>>>>>>>> From: Jason Villalta [mailto:jason@rubixnet.com]
> >>>>>>>>>> Sent: Thursday, February 14, 2013 1:45 PM
> >>>>>>>>>> To: cloudstack-users@incubator.apache.org
> >>>>>>>>>> Subject: Change qcow disk cache=none
> >>>>>>>>>>
> >>>>>>>>>> Is there a way to change this kvm disk cache setting in
> >>>>>>>>>> Cloudstack or
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> template?
> >>>>>>>>>>
> >>>>>>>>>>
> >>>>>>>>>> If so where is the setting applied?
> >>>>>>>>>>
> >>>>>>>>>> --
> >>>>>>>>>> --
> >>>>>>>>>> Jason Villalta
> >>>>>>>>>> Co-founder
> >>>>>>>>>> 800.799.4407 | www.RubixTechnology.com
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> --
> >>>>>>>> --
> >>>>>>>> Jason Villalta
> >>>>>>>> Co-founder
> >>>>>>>> 800.799.4407 | www.RubixTechnology.com
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>
> >>
> >>
> >
>
>
>
> --
> --
> Jason Villalta
> Co-founder
> 800.799.4407 | www.RubixTechnology.com
>

--f46d043d64b345ac2a04d5c5526a--