Hi Dinesh,
many thanks for your time, the expertise you've kindly shared on this
discussion.
I believe that Santosh has volunteered ;) to provide some text on the
firewall interaction. Any other contributions are welcome and greatly
appreciated.
Regards,
Greg
On Wed, Oct 23, 2019 at 3:54 PM
You're welcome Greg. I'm glad my input was useful,
Dinesh
On Oct 24, 2019, 1:33 AM +0530, Greg Mirsky , wrote:
> Hi Dinesh,
> many thanks for your time, the expertise you've kindly shared on this
> discussion.
> I believe that Santosh has volunteered ;) to provide some text on the
> firewall
Looks good to me Greg. I see that the text around the use of the inner
IP address as also quite acceptable. Will you add any words about the
firewall?
Dinesh
On Wed, Oct 23, 2019 at 8:36 PM, Greg Mirsky
wrote:
Hi Dinesh, et al.,
please check the updated version that removed the reference
Thanks Joel.
I see the issue. In the case of IRB, the VTEP will likely have IP
addresses assigned from the tenant space for each VNI. But if there is no
IRB, then it could be a problem. Thus far, my assumption had been that the
underlay address would be used and that the inner addresses would
Yes, what I discussed with Anoop was on Greg's option #3.
Respecting BFD over VxLAN, option #2 and #3 both are ok to me, I have no
preference.
Respecting BFD over Geneve, option #2 and #3 both are ok to me, although I
personally prefer #3.
Best Regards,
Xiao Min
原始邮件
Greg,
I think the draft is fine as is.
I discussion with Xiao Min was about #3 and I see that as unnecessary until
we have a draft that explains why that is needed in the context of the NVO3
architecture.
Anoop
On Tue, Oct 22, 2019 at 11:17 AM Greg Mirsky wrote:
> Hi Anoop, et al.,
> I agree
Dinesh,
Please see my inline comments [SPK]
>
> - In section 3, there's a sentence that is: "BFD packets intended for a
> Hypervisor VTEP MUST NOT..". I recommend getting rid of the word
> "Hypervisor" ashe logic applies to any VTEP.
>
> [SPK] Thanks for comments. We will change this.
> -
Anoop,
I guess there were multiple discussion over this should we have inner
TTL as 1 or destination IP address as 127/8 range so that if packet gets
exposed in underlay it should not be routed via underlay to VTEP.
Thanks
Santosh P K
On Wed, Oct 23, 2019 at 11:40 AM Anoop Ghanwani
wrote:
>
Anoop, you refer to "the destination VTEP's IP address". Since this is
a field inside the Ethernet header inside the VxLAN header, what VTEP
assigned IP address? The customer (whose address space this is in may
not be using IP. Or may be using IP and presumably has NOT assigned an
IP
Hi Greg,
The part about the use of 127/8 address appears to be a new thing
introduced in the version of the draft that is as of yet unpublished. What
was the motivation for the change? Previously, the DA was simply set to
the destination VTEP's IP address which seemed fine.
Anoop
On Tue, Oct
I have the same feeling as Anoop. Greg, can you please point me to the
latest draft so that I can quickly glance through it to be doubly sure,
Dinesh
On Wed, Oct 23, 2019 at 4:35 AM, Anoop Ghanwani
wrote:
Greg,
I think the draft is fine as is.
I discussion with Xiao Min was about #3 and I
Greg,
Two comments, one minor and one maybe not.
- In section 3, there's a sentence that is: "BFD packets intended for a
Hypervisor VTEP MUST NOT..". I recommend getting rid of the word
"Hypervisor" ashe logic applies to any VTEP.
- You already explained the precedence of the use of 127/8
12 matches
Mail list logo
