Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: OpenStack: Operators

NFS Implementation Details

 

 

OpenStack operators RSS feed   Index | Next | Previous | View Threaded


joe.topjian at cybera

Jun 9, 2012, 8:37 AM

Post #1 of 8 (1067 views)
Permalink
NFS Implementation Details

Hello,

I did not want to diverge too far off topic with the current Live Migration
thread.

I'm wondering if anyone would mind sharing their NFS implementation
details?

For example: have you settled on a ratio of NFS Servers to Compute Nodes?
Are you using NFS service on Linux or a commercial filer such as a NetApp
appliance?

Diego mentioned using NFS for all aspects of OpenStack - including volumes.
I was wondering if you could explain that a little bit more?

Thanks,
Joe

--
Joe Topjian
Systems Administrator
Cybera Inc.

www.cybera.ca

Cybera is a not-for-profit organization that works to spur and support
innovation, for the economic benefit of Alberta, through the use
of cyberinfrastructure.


igor.laskovy at gmail

Jun 10, 2012, 2:29 AM

Post #2 of 8 (1007 views)
Permalink
Re: NFS Implementation Details [In reply to]

+1 to this request! This is very interested to me too.

On Sat, Jun 9, 2012 at 6:37 PM, Joe Topjian <joe.topjian [at] cybera> wrote:
> Hello,
>
> I did not want to diverge too far off topic with the current Live Migration
> thread.
>
> I'm wondering if anyone would mind sharing their NFS implementation
> details?
>
> For example: have you settled on a ratio of NFS Servers to Compute Nodes?
> Are you using NFS service on Linux or a commercial filer such as a NetApp
> appliance?
>
> Diego mentioned using NFS for all aspects of OpenStack - including volumes.
> I was wondering if you could explain that a little bit more?
>
> Thanks,
> Joe
>
> --
> Joe Topjian
> Systems Administrator
> Cybera Inc.
>
> www.cybera.ca
>
> Cybera is a not-for-profit organization that works to spur and support
> innovation, for the economic benefit of Alberta, through the use
> of cyberinfrastructure.
>
>
> _______________________________________________
> Openstack-operators mailing list
> Openstack-operators [at] lists
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>



--
Igor Laskovy
Kiev, Ukraine
_______________________________________________
Openstack-operators mailing list
Openstack-operators [at] lists
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


kraemer at avarteq

Jun 11, 2012, 4:28 AM

Post #3 of 8 (990 views)
Permalink
Re: NFS Implementation Details [In reply to]

Hey Guys,

so i'm using NFS with OpenStack.
I see two solutions, the cloud way and the "shared services" way :).

Cloudway:
* Attach a volume to your NFS instance.
* Mount the attached volume to /var/exports/ or where ever you want. But
you need to configure this in NFS.
* Export you directories to the various fixed IP's of your internal network

my instances automatically. It's no problem then if the instance go in
error state. Maybe a reboot of the instance helps but if not you need an
automation.


Shared services:
We use a couple of shared services like DB-Clusters (postgres, mysql,
mongodb), rabbitmq, mailserver and NFS.
But this way is more complex because you need static routes on each
compute node. Melange can configure static routes but i think it's to
oversized to use melange only for routing. So changed the code of the
network manager to configure my routes. It's not a big deal beacuase i
need only 2 static routes ;)

I hope this short answer helps you a little bit or shows an other way :>


Greetings
Heiko

Am 10.06.2012 11:29, schrieb Igor Laskovy:
> +1 to this request! This is very interested to me too.
>
> On Sat, Jun 9, 2012 at 6:37 PM, Joe Topjian <joe.topjian [at] cybera> wrote:
>> Hello,
>>
>> I did not want to diverge too far off topic with the current Live Migration
>> thread.
>>
>> I'm wondering if anyone would mind sharing their NFS implementation
>> details?
>>
>> For example: have you settled on a ratio of NFS Servers to Compute Nodes?
>> Are you using NFS service on Linux or a commercial filer such as a NetApp
>> appliance?
>>
>> Diego mentioned using NFS for all aspects of OpenStack - including volumes.
>> I was wondering if you could explain that a little bit more?
>>
>> Thanks,
>> Joe
>>
>> --
>> Joe Topjian
>> Systems Administrator
>> Cybera Inc.
>>
>> www.cybera.ca
>>
>> Cybera is a not-for-profit organization that works to spur and support
>> innovation, for the economic benefit of Alberta, through the use
>> of cyberinfrastructure.
>>
>>
>> _______________________________________________
>> Openstack-operators mailing list
>> Openstack-operators [at] lists
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>
>
>

--
B. Sc. Informatik
Heiko Krämer
CIO/Administrator

Twitter: @railshoster
Avarteq GmbH
Zweigstelle:
Prinzessinnenstr. 20, 10969 Berlin

----
Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian Fischer
Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.: DE262633168
Sitz:
Science Park 2
66123 Saarbrücken

Tel: +49 (0)681 / 309 64 190
Fax: +49 (0)681 / 309 64 191

Visit:
http://www.enterprise-rails.de/
Attachments: kraemer.vcf (0.30 KB)
  signature.asc (0.54 KB)


igor.laskovy at gmail

Jun 11, 2012, 4:53 AM

Post #4 of 8 (993 views)
Permalink
Re: NFS Implementation Details [In reply to]

Hello Heiko,

Totally confused about your "shared service" way))

In cloud way, are you using any clustered NFS solution? How big it?

Igor Laskovy
Kiev, Ukraine
On Jun 11, 2012 2:28 PM, "Heiko Krämer" <kraemer [at] avarteq> wrote:

> Hey Guys,
>
> so i'm using NFS with OpenStack.
> I see two solutions, the cloud way and the "shared services" way :).
>
> Cloudway:
> * Attach a volume to your NFS instance.
> * Mount the attached volume to /var/exports/ or where ever you want. But
> you need to configure this in NFS.
> * Export you directories to the various fixed IP's of your internal network
>
> => Instance failes ..... so i'm using chef to install and configure all
> my instances automatically. It's no problem then if the instance go in
> error state. Maybe a reboot of the instance helps but if not you need an
> automation.
>
>
> Shared services:
> We use a couple of shared services like DB-Clusters (postgres, mysql,
> mongodb), rabbitmq, mailserver and NFS.
> But this way is more complex because you need static routes on each
> compute node. Melange can configure static routes but i think it's to
> oversized to use melange only for routing. So changed the code of the
> network manager to configure my routes. It's not a big deal beacuase i
> need only 2 static routes ;)
>
> I hope this short answer helps you a little bit or shows an other way :>
>
>
> Greetings
> Heiko
>
> Am 10.06.2012 11:29, schrieb Igor Laskovy:
> > +1 to this request! This is very interested to me too.
> >
> > On Sat, Jun 9, 2012 at 6:37 PM, Joe Topjian <joe.topjian [at] cybera>
> wrote:
> >> Hello,
> >>
> >> I did not want to diverge too far off topic with the current Live
> Migration
> >> thread.
> >>
> >> I'm wondering if anyone would mind sharing their NFS implementation
> >> details?
> >>
> >> For example: have you settled on a ratio of NFS Servers to Compute
> Nodes?
> >> Are you using NFS service on Linux or a commercial filer such as a
> NetApp
> >> appliance?
> >>
> >> Diego mentioned using NFS for all aspects of OpenStack - including
> volumes.
> >> I was wondering if you could explain that a little bit more?
> >>
> >> Thanks,
> >> Joe
> >>
> >> --
> >> Joe Topjian
> >> Systems Administrator
> >> Cybera Inc.
> >>
> >> www.cybera.ca
> >>
> >> Cybera is a not-for-profit organization that works to spur and support
> >> innovation, for the economic benefit of Alberta, through the use
> >> of cyberinfrastructure.
> >>
> >>
> >> _______________________________________________
> >> Openstack-operators mailing list
> >> Openstack-operators [at] lists
> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
> >>
> >
> >
>
> --
> B. Sc. Informatik
> Heiko Krämer
> CIO/Administrator
>
> Twitter: @railshoster
> Avarteq GmbH
> Zweigstelle:
> Prinzessinnenstr. 20, 10969 Berlin
>
> ----
> Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian Fischer
> Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.: DE262633168
> Sitz:
> Science Park 2
> 66123 Saarbrücken
>
> Tel: +49 (0)681 / 309 64 190
> Fax: +49 (0)681 / 309 64 191
>
> Visit:
> http://www.enterprise-rails.de/
>
>
> _______________________________________________
> Openstack-operators mailing list
> Openstack-operators [at] lists
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>


joe.topjian at cybera

Jun 11, 2012, 7:01 AM

Post #5 of 8 (994 views)
Permalink
Re: NFS Implementation Details [In reply to]

Hi Heiko,

Thank you for the information.

With regard to your cloudway, do you find that you get acceptable transfer
rates given your NFS service is hosted in an instance - or do you see a
noticeable decrease in performance due to virtualization overhead?

How many instances are using your cloud NFS server?

Thanks again,
Joe


On Mon, Jun 11, 2012 at 5:28 AM, Heiko Krämer <kraemer [at] avarteq> wrote:

> Hey Guys,
>
> so i'm using NFS with OpenStack.
> I see two solutions, the cloud way and the "shared services" way :).
>
> Cloudway:
> * Attach a volume to your NFS instance.
> * Mount the attached volume to /var/exports/ or where ever you want. But
> you need to configure this in NFS.
> * Export you directories to the various fixed IP's of your internal network
>
> => Instance failes ..... so i'm using chef to install and configure all
> my instances automatically. It's no problem then if the instance go in
> error state. Maybe a reboot of the instance helps but if not you need an
> automation.
>
>
> Shared services:
> We use a couple of shared services like DB-Clusters (postgres, mysql,
> mongodb), rabbitmq, mailserver and NFS.
> But this way is more complex because you need static routes on each
> compute node. Melange can configure static routes but i think it's to
> oversized to use melange only for routing. So changed the code of the
> network manager to configure my routes. It's not a big deal beacuase i
> need only 2 static routes ;)
>
> I hope this short answer helps you a little bit or shows an other way :>
>
>
> Greetings
> Heiko
>
> Am 10.06.2012 11:29, schrieb Igor Laskovy:
> > +1 to this request! This is very interested to me too.
> >
> > On Sat, Jun 9, 2012 at 6:37 PM, Joe Topjian <joe.topjian [at] cybera>
> wrote:
> >> Hello,
> >>
> >> I did not want to diverge too far off topic with the current Live
> Migration
> >> thread.
> >>
> >> I'm wondering if anyone would mind sharing their NFS implementation
> >> details?
> >>
> >> For example: have you settled on a ratio of NFS Servers to Compute
> Nodes?
> >> Are you using NFS service on Linux or a commercial filer such as a
> NetApp
> >> appliance?
> >>
> >> Diego mentioned using NFS for all aspects of OpenStack - including
> volumes.
> >> I was wondering if you could explain that a little bit more?
> >>
> >> Thanks,
> >> Joe
> >>
> >> --
> >> Joe Topjian
> >> Systems Administrator
> >> Cybera Inc.
> >>
> >> www.cybera.ca
> >>
> >> Cybera is a not-for-profit organization that works to spur and support
> >> innovation, for the economic benefit of Alberta, through the use
> >> of cyberinfrastructure.
> >>
> >>
> >> _______________________________________________
> >> Openstack-operators mailing list
> >> Openstack-operators [at] lists
> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
> >>
> >
> >
>
> --
> B. Sc. Informatik
> Heiko Krämer
> CIO/Administrator
>
> Twitter: @railshoster
> Avarteq GmbH
> Zweigstelle:
> Prinzessinnenstr. 20, 10969 Berlin
>
> ----
> Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian Fischer
> Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.: DE262633168
> Sitz:
> Science Park 2
> 66123 Saarbrücken
>
> Tel: +49 (0)681 / 309 64 190
> Fax: +49 (0)681 / 309 64 191
>
> Visit:
> http://www.enterprise-rails.de/
>
>
> _______________________________________________
> Openstack-operators mailing list
> Openstack-operators [at] lists
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>


--
Joe Topjian
Systems Administrator
Cybera Inc.

www.cybera.ca

Cybera is a not-for-profit organization that works to spur and support
innovation, for the economic benefit of Alberta, through the use
of cyberinfrastructure.


kraemer at avarteq

Jun 14, 2012, 11:35 PM

Post #6 of 8 (976 views)
Permalink
Re: NFS Implementation Details [In reply to]

Hey Joe,

Am Montag, den 11.06.2012, 08:01 -0600 schrieb Joe Topjian:
> Hi Heiko,
>
>
> Thank you for the information.
>
>
> With regard to your cloudway, do you find that you get acceptable
> transfer rates given your NFS service is hosted in an instance - or do
> you see a noticeable decrease in performance due to virtualization
> overhead?
>
hmmm not good but acceptable. I'm using kvm atm with virtio drivers and
the transfer rate is ok. I think you shold decide matching the
respective application pupose if the transfer rate enough or not.


>
> How many instances are using your cloud NFS server?
>
It's a cluster of servers. I'm using glusterfs and a loadbalancing with
write protection.
I can't get an answer of this question for "one" server. You have to
many facts (like lan throughput, hdd capacity, hdd performance and so
on) to answer this.

Cheers
Heiko
>
> Thanks again,
> Joe
>
>
> On Mon, Jun 11, 2012 at 5:28 AM, Heiko Krämer <kraemer [at] avarteq>
> wrote:
> Hey Guys,
>
> so i'm using NFS with OpenStack.
> I see two solutions, the cloud way and the "shared services"
> way :).
>
> Cloudway:
> * Attach a volume to your NFS instance.
> * Mount the attached volume to /var/exports/ or where ever you
> want. But
> you need to configure this in NFS.
> * Export you directories to the various fixed IP's of your
> internal network
>
> => Instance failes ..... so i'm using chef to install and
> configure all
> my instances automatically. It's no problem then if the
> instance go in
> error state. Maybe a reboot of the instance helps but if not
> you need an
> automation.
>
>
> Shared services:
> We use a couple of shared services like DB-Clusters (postgres,
> mysql,
> mongodb), rabbitmq, mailserver and NFS.
> But this way is more complex because you need static routes on
> each
> compute node. Melange can configure static routes but i think
> it's to
> oversized to use melange only for routing. So changed the code
> of the
> network manager to configure my routes. It's not a big deal
> beacuase i
> need only 2 static routes ;)
>
> I hope this short answer helps you a little bit or shows an
> other way :>
>
>
> Greetings
> Heiko
>
> Am 10.06.2012 11:29, schrieb Igor Laskovy:
> > +1 to this request! This is very interested to me too.
> >
> > On Sat, Jun 9, 2012 at 6:37 PM, Joe Topjian
> <joe.topjian [at] cybera> wrote:
> >> Hello,
> >>
> >> I did not want to diverge too far off topic with the
> current Live Migration
> >> thread.
> >>
> >> I'm wondering if anyone would mind sharing their NFS
> implementation
> >> details?
> >>
> >> For example: have you settled on a ratio of NFS Servers to
> Compute Nodes?
> >> Are you using NFS service on Linux or a commercial filer
> such as a NetApp
> >> appliance?
> >>
> >> Diego mentioned using NFS for all aspects of OpenStack -
> including volumes.
> >> I was wondering if you could explain that a little bit
> more?
> >>
> >> Thanks,
> >> Joe
> >>
> >> --
> >> Joe Topjian
> >> Systems Administrator
> >> Cybera Inc.
> >>
> >> www.cybera.ca
> >>
> >> Cybera is a not-for-profit organization that works to spur
> and support
> >> innovation, for the economic benefit of Alberta, through
> the use
> >> of cyberinfrastructure.
> >>
> >>
> >> _______________________________________________
> >> Openstack-operators mailing list
> >> Openstack-operators [at] lists
> >>
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
> >>
> >
> >
>
> --
>
> B. Sc. Informatik
> Heiko Krämer
> CIO/Administrator
>
> Twitter: @railshoster
> Avarteq GmbH
> Zweigstelle:
> Prinzessinnenstr. 20, 10969 Berlin
>
> ----
> Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian
> Fischer
> Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.:
> DE262633168
> Sitz:
> Science Park 2
> 66123 Saarbrücken
>
> Tel: +49 (0)681 / 309 64 190
> Fax: +49 (0)681 / 309 64 191
>
> Visit:
> http://www.enterprise-rails.de/
>
>
> _______________________________________________
> Openstack-operators mailing list
> Openstack-operators [at] lists
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>
>
>
>
>
> --
> Joe Topjian
> Systems Administrator
> Cybera Inc.
>
>
> www.cybera.ca
>
>
> Cybera is a not-for-profit organization that works to spur and support
> innovation, for the economic benefit of Alberta, through the use
> of cyberinfrastructure.
>
>

--
B. Sc. Informatik
Heiko Krämer
CIO/Administrator

Twitter: @railshoster
Avarteq GmbH
Zweigstelle:
Prinzessinnenstr. 20, 10969 Berlin

----
Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian Fischer
Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.: DE262633168
Sitz:
Science Park 2
66123 Saarbrücken

Tel: +49 (0)681 / 309 64 190
Fax: +49 (0)681 / 309 64 191

Visit:
http://www.enterprise-rails.de/

_______________________________________________
Openstack-operators mailing list
Openstack-operators [at] lists
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


kraemer at avarteq

Jun 14, 2012, 11:40 PM

Post #7 of 8 (990 views)
Permalink
Re: NFS Implementation Details [In reply to]

Hey Igor,

Am Montag, den 11.06.2012, 14:53 +0300 schrieb Igor Laskovy:
> Hello Heiko,
>
> Totally confused about your "shared service" way))
If you went to share an service with more then one project, in this case
NFS service, you need to create static routes. All networks must have
the chance to get access to this service. On this way you can share
different services like rabbitMQ-cluster, mysql-cluster, postgres and so
on with your customers.

So you don't have an overhead of instances in you cloud.
> In cloud way, are you using any clustered NFS solution? How big it?
See my answer to Joe
>
Cheers from Berlin,
Heiko
> Igor Laskovy
> Kiev, Ukraine
>
> On Jun 11, 2012 2:28 PM, "Heiko Krämer" <kraemer [at] avarteq> wrote:
> Hey Guys,
>
> so i'm using NFS with OpenStack.
> I see two solutions, the cloud way and the "shared services"
> way :).
>
> Cloudway:
> * Attach a volume to your NFS instance.
> * Mount the attached volume to /var/exports/ or where ever you
> want. But
> you need to configure this in NFS.
> * Export you directories to the various fixed IP's of your
> internal network
>
> => Instance failes ..... so i'm using chef to install and
> configure all
> my instances automatically. It's no problem then if the
> instance go in
> error state. Maybe a reboot of the instance helps but if not
> you need an
> automation.
>
>
> Shared services:
> We use a couple of shared services like DB-Clusters (postgres,
> mysql,
> mongodb), rabbitmq, mailserver and NFS.
> But this way is more complex because you need static routes on
> each
> compute node. Melange can configure static routes but i think
> it's to
> oversized to use melange only for routing. So changed the code
> of the
> network manager to configure my routes. It's not a big deal
> beacuase i
> need only 2 static routes ;)
>
> I hope this short answer helps you a little bit or shows an
> other way :>
>
>
> Greetings
> Heiko
>
> Am 10.06.2012 11:29, schrieb Igor Laskovy:
> > +1 to this request! This is very interested to me too.
> >
> > On Sat, Jun 9, 2012 at 6:37 PM, Joe Topjian
> <joe.topjian [at] cybera> wrote:
> >> Hello,
> >>
> >> I did not want to diverge too far off topic with the
> current Live Migration
> >> thread.
> >>
> >> I'm wondering if anyone would mind sharing their NFS
> implementation
> >> details?
> >>
> >> For example: have you settled on a ratio of NFS Servers to
> Compute Nodes?
> >> Are you using NFS service on Linux or a commercial filer
> such as a NetApp
> >> appliance?
> >>
> >> Diego mentioned using NFS for all aspects of OpenStack -
> including volumes.
> >> I was wondering if you could explain that a little bit
> more?
> >>
> >> Thanks,
> >> Joe
> >>
> >> --
> >> Joe Topjian
> >> Systems Administrator
> >> Cybera Inc.
> >>
> >> www.cybera.ca
> >>
> >> Cybera is a not-for-profit organization that works to spur
> and support
> >> innovation, for the economic benefit of Alberta, through
> the use
> >> of cyberinfrastructure.
> >>
> >>
> >> _______________________________________________
> >> Openstack-operators mailing list
> >> Openstack-operators [at] lists
> >>
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
> >>
> >
> >
>
> --
> B. Sc. Informatik
> Heiko Krämer
> CIO/Administrator
>
> Twitter: @railshoster
> Avarteq GmbH
> Zweigstelle:
> Prinzessinnenstr. 20, 10969 Berlin
>
> ----
> Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian
> Fischer
> Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.:
> DE262633168
> Sitz:
> Science Park 2
> 66123 Saarbrücken
>
> Tel: +49 (0)681 / 309 64 190
> Fax: +49 (0)681 / 309 64 191
>
> Visit:
> http://www.enterprise-rails.de/
>
>
> _______________________________________________
> Openstack-operators mailing list
> Openstack-operators [at] lists
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>

--
B. Sc. Informatik
Heiko Krämer
CIO/Administrator

Twitter: @railshoster
Avarteq GmbH
Zweigstelle:
Prinzessinnenstr. 20, 10969 Berlin

----
Geschäftsführer: Alexander Faißt, Dipl.-Inf.(FH) Julian Fischer
Handelsregister: AG Saarbrücken HRB 17413, Ust-IdNr.: DE262633168
Sitz:
Science Park 2
66123 Saarbrücken

Tel: +49 (0)681 / 309 64 190
Fax: +49 (0)681 / 309 64 191

Visit:
http://www.enterprise-rails.de/

_______________________________________________
Openstack-operators mailing list
Openstack-operators [at] lists
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


trapni at gmail

Jun 15, 2012, 1:20 AM

Post #8 of 8 (977 views)
Permalink
Re: NFS Implementation Details [In reply to]

Hey ALL,

On Fri, Jun 15, 2012 at 8:35 AM, Heiko Krämer <kraemer [at] avarteq> wrote:

> Hey Joe,
>
> Am Montag, den 11.06.2012, 08:01 -0600 schrieb Joe Topjian:
> > Hi Heiko,
>
[....]

> > With regard to your cloudway, do you find that you get acceptable
> > transfer rates given your NFS service is hosted in an instance - or do
> > you see a noticeable decrease in performance due to virtualization
> > overhead?
> >
> hmmm not good but acceptable. I'm using kvm atm with virtio drivers and
> the transfer rate is ok. I think you shold decide matching the
> respective application pupose if the transfer rate enough or not.


I checked, and it seems like on Ubuntu 12.04 LTS virtio is used by default,
which is good. :)
However, what services are you running in VMs?

I am planning to put mySQL (first testing with slaves) nodes into KVM
and compare performance with the current bare-metal SSD-powered
equivalents.

>
> > How many instances are using your cloud NFS server?
> >
> It's a cluster of servers. I'm using glusterfs and a loadbalancing with
> write protection.
>

Would you mind in giving a little more detail on this one?
* How many glusterfs nodes do you have?
* Are they dedicated (no other services running on it)?
* How exactly do you connect the clients (nova-compute nodes) with it?
glusterfs.fuse or NFS?
* wrt last question, what about fail-over IP in the latter case?
* What kind of glusterfs topology do you have? (replica=N, distributed,
distributed replicated?)
* Are you using FibreChannel/SAN for Glusterfs I/O ? (Is it worth the cost?)

I can't get an answer of this question for "one" server. You have to
> many facts (like lan throughput, hdd capacity, hdd performance and so
> on) to answer this.
>

Are you using SSD or rotated media for /var/lib/nova/instances?

Cheers from the other side of Berlin.
Christian.

OpenStack operators RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.