You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@cloudstack.apache.org by Andrija Panic <an...@gmail.com> on 2015/05/03 13:49:16 UTC
[arhitecture help] Management vs. Secondary network connectivity dependencies
Hi guys,
I was wondering if someone experienced would be kind to share their opinion
on folowing subject - since I'm 95% clear with design/ho stuff works, but
again missing some info - after I have done my RTFM many times, and I had
also check in details ShapeBlue's diagrams on Networking Deep Dive etc.
Subject: Management vs Secondary Storage connectivity dependencies
Current setup:
- We start with 1 POD, 1 Cluster
- All KVM hosts are directly attached to mgmt, primary storage, and
secondary storage networks/vlans (via 3 dedicated NICs/vlans - each one on
a separate/non-overlapping /16 network), so no rouing required for any KVM
host to reach any resource on all of these 3 networks...
- Zone-wide Primary Storage (CEPH) - again, directly attached to all KVM
hosts - this means all future PODs will be also directly attached to
CEPH/Primary Storage Net - no routing...
- NFS Zone-Wide Secondary Storage (of course)
For Primary Storage traffic/network I dont have any questions - I
understand how stuff works (thx for latest docs and ShapeBlue blogs) -
except possible zone-wide vs cluster-wide discussion...
Questions that I have - Secondary Storage network...:
1. Since all KVM hosts are directly attached (via dedicated interface)
to both MGMT and SEC storage network - is there any routing that needs to
be done on the L3 switch ? My understanding not, but some confirmation
would be nice please.
2. What about when we have multiple PODs/Clusters - does it make any
difference versus 1 single POD/Cluster? I understand there needs to be
routing connectivity over L3 switch between different Management networks
of each POD, but what about storage? I expect I would need to attach/trunk
Secondary Storage Net to all KVM hosts in all PODs... ?
3. What other networks (except MGMT) does Management server needs to be
connected to ? I understand there needs to be eather routed or dirrect
(dedicated NIC) connection to Secondary Storage network - at least during
ACS deployment, during sys template preseed on NFS ?
4. Is Secondary Storage network required for anything else on MGMT
servers, after the ACS template preseed/provisioning ? I understand not,
but again would benefit some confirmation please.
5. Finally - in terms of pure ACS aspect - is there a difference between
KVM hosts being connected directly to Secondary Stor Net (via dedicated
NIC/trunked) or using routing - in other words, can we say that something
like folowing is true: If the KVM host doesnt have direct attached
network to it self (primary, secondary) = it will used Management Net
gateway to try to reach its destination (Pimary/Secondary storage for
example) meaning it will use routing for this.
6. What is really Secondary Storage Network Gateway used for in ACS ?
I apologize for having long questions - I just try to understand
connectivity dependencies between Sec and Mgmt networks and will be really
thankfull for clarifying these...
Regards,
--
Andrija Panić
Re: [arhitecture help] Management vs. Secondary network connectivity dependencies
Posted by Andrija Panic <an...@gmail.com>.
anyone :)
On 3 May 2015 at 13:49, Andrija Panic <an...@gmail.com> wrote:
> Hi guys,
>
> I was wondering if someone experienced would be kind to share their
> opinion on folowing subject - since I'm 95% clear with design/ho stuff
> works, but again missing some info - after I have done my RTFM many times,
> and I had also check in details ShapeBlue's diagrams on Networking Deep
> Dive etc.
>
> Subject: Management vs Secondary Storage connectivity dependencies
>
>
> Current setup:
>
> - We start with 1 POD, 1 Cluster
> - All KVM hosts are directly attached to mgmt, primary storage, and
> secondary storage networks/vlans (via 3 dedicated NICs/vlans - each one on
> a separate/non-overlapping /16 network), so no rouing required for any KVM
> host to reach any resource on all of these 3 networks...
> - Zone-wide Primary Storage (CEPH) - again, directly attached to all KVM
> hosts - this means all future PODs will be also directly attached to
> CEPH/Primary Storage Net - no routing...
> - NFS Zone-Wide Secondary Storage (of course)
>
>
> For Primary Storage traffic/network I dont have any questions - I
> understand how stuff works (thx for latest docs and ShapeBlue blogs) -
> except possible zone-wide vs cluster-wide discussion...
>
>
>
> Questions that I have - Secondary Storage network...:
>
>
> 1. Since all KVM hosts are directly attached (via dedicated
> interface) to both MGMT and SEC storage network - is there any routing that
> needs to be done on the L3 switch ? My understanding not, but some
> confirmation would be nice please.
> 2. What about when we have multiple PODs/Clusters - does it make any
> difference versus 1 single POD/Cluster? I understand there needs to be
> routing connectivity over L3 switch between different Management networks
> of each POD, but what about storage? I expect I would need to attach/trunk
> Secondary Storage Net to all KVM hosts in all PODs... ?
> 3. What other networks (except MGMT) does Management server needs to
> be connected to ? I understand there needs to be eather routed or dirrect
> (dedicated NIC) connection to Secondary Storage network - at least during
> ACS deployment, during sys template preseed on NFS ?
> 4. Is Secondary Storage network required for anything else on MGMT
> servers, after the ACS template preseed/provisioning ? I understand not,
> but again would benefit some confirmation please.
> 5. Finally - in terms of pure ACS aspect - is there a difference
> between KVM hosts being connected directly to Secondary Stor Net (via
> dedicated NIC/trunked) or using routing - in other words, can we say that
> something like folowing is true: If the KVM host doesnt have direct
> attached network to it self (primary, secondary) = it will used Management
> Net gateway to try to reach its destination (Pimary/Secondary storage for
> example) meaning it will use routing for this.
> 6. What is really Secondary Storage Network Gateway used for in ACS ?
>
>
> I apologize for having long questions - I just try to understand
> connectivity dependencies between Sec and Mgmt networks and will be really
> thankfull for clarifying these...
>
> Regards,
> --
>
> Andrija Panić
>
--
Andrija Panić
Re: [arhitecture help] Management vs. Secondary network connectivity dependencies
Posted by Andrija Panic <an...@gmail.com>.
anyone :)
On 3 May 2015 at 13:49, Andrija Panic <an...@gmail.com> wrote:
> Hi guys,
>
> I was wondering if someone experienced would be kind to share their
> opinion on folowing subject - since I'm 95% clear with design/ho stuff
> works, but again missing some info - after I have done my RTFM many times,
> and I had also check in details ShapeBlue's diagrams on Networking Deep
> Dive etc.
>
> Subject: Management vs Secondary Storage connectivity dependencies
>
>
> Current setup:
>
> - We start with 1 POD, 1 Cluster
> - All KVM hosts are directly attached to mgmt, primary storage, and
> secondary storage networks/vlans (via 3 dedicated NICs/vlans - each one on
> a separate/non-overlapping /16 network), so no rouing required for any KVM
> host to reach any resource on all of these 3 networks...
> - Zone-wide Primary Storage (CEPH) - again, directly attached to all KVM
> hosts - this means all future PODs will be also directly attached to
> CEPH/Primary Storage Net - no routing...
> - NFS Zone-Wide Secondary Storage (of course)
>
>
> For Primary Storage traffic/network I dont have any questions - I
> understand how stuff works (thx for latest docs and ShapeBlue blogs) -
> except possible zone-wide vs cluster-wide discussion...
>
>
>
> Questions that I have - Secondary Storage network...:
>
>
> 1. Since all KVM hosts are directly attached (via dedicated
> interface) to both MGMT and SEC storage network - is there any routing that
> needs to be done on the L3 switch ? My understanding not, but some
> confirmation would be nice please.
> 2. What about when we have multiple PODs/Clusters - does it make any
> difference versus 1 single POD/Cluster? I understand there needs to be
> routing connectivity over L3 switch between different Management networks
> of each POD, but what about storage? I expect I would need to attach/trunk
> Secondary Storage Net to all KVM hosts in all PODs... ?
> 3. What other networks (except MGMT) does Management server needs to
> be connected to ? I understand there needs to be eather routed or dirrect
> (dedicated NIC) connection to Secondary Storage network - at least during
> ACS deployment, during sys template preseed on NFS ?
> 4. Is Secondary Storage network required for anything else on MGMT
> servers, after the ACS template preseed/provisioning ? I understand not,
> but again would benefit some confirmation please.
> 5. Finally - in terms of pure ACS aspect - is there a difference
> between KVM hosts being connected directly to Secondary Stor Net (via
> dedicated NIC/trunked) or using routing - in other words, can we say that
> something like folowing is true: If the KVM host doesnt have direct
> attached network to it self (primary, secondary) = it will used Management
> Net gateway to try to reach its destination (Pimary/Secondary storage for
> example) meaning it will use routing for this.
> 6. What is really Secondary Storage Network Gateway used for in ACS ?
>
>
> I apologize for having long questions - I just try to understand
> connectivity dependencies between Sec and Mgmt networks and will be really
> thankfull for clarifying these...
>
> Regards,
> --
>
> Andrija Panić
>
--
Andrija Panić