You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cloudstack.apache.org by Nux! <nu...@li.nux.ro> on 2017/11/20 16:08:50 UTC
Circumventing VXLAN MTU issues
Hello,
I am playing around with the native VXLAN implementation and I have of course hit the situation where MTU on the host gets chomped by 50 bytes and traffic crawls to a stop in the VMs.
How are you people circumventing this? The easiest, but kind of inconvenient for customers would be to set MTU 1450 inside the VM, though it could be baked into the templates or via cloud-init.
Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I understand ovs can dynamically adjust MTUs to compensate (?), but at the same time this VXLAN implementation suffers from some limitations (multicast etc).
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Andrija Panic <an...@gmail.com>.
@devs any plan to implement VXLAN as isolation for also Private Gateway
functionality ?
On 21 December 2017 at 15:28, Andrija Panic <an...@gmail.com> wrote:
> while there, here is another one (just below that one) :) which I learned
> the very hard way, after 2 years in production (when clients started
> complaining on random traffic drop) : http://docs.cloudstack.
> apache.org/en/latest/networking/vxlan.html#important-note-on-max-number-
> of-multicast-groups-and-thus-vxlan-intefaces
>
> Very sneaky and ugly: Ubuntu kernel 3.16 = random traffic drops, but all
> vxlan interfaces are UP etc..., Ubuntu kernel 4.x vxlan interface would NOT
> come up AT ALL, so a hard drop of packets, and there is where we saw the
> issue...
>
> Cheers
>
>
> On 21 December 2017 at 15:13, Nux! <nu...@li.nux.ro> wrote:
>
>> Thanks Andrija, well done, I have indeed read your document.
>> Using MTU 9000 solved my problem. :)
>>
>> Lucian
>>
>> --
>> Sent from the Delta quadrant using Borg technology!
>>
>> Nux!
>> www.nux.ro
>>
>> ----- Original Message -----
>> > From: "Andrija Panic" <an...@gmail.com>
>> > To: "dev" <de...@cloudstack.apache.org>
>> > Cc: "users" <us...@cloudstack.apache.org>
>> > Sent: Thursday, 21 December, 2017 14:09:43
>> > Subject: Re: Circumventing VXLAN MTU issues
>>
>> > Hi Nux,
>> >
>> > there is one rare contributions from my side :D to the ACS
>> (documentation,
>> > 2.5 years ago) - check it here:
>> > http://docs.cloudstack.apache.org/en/latest/networking/vxlan
>> .html#important-note-on-mtu-size
>> >
>> > We are using it extensively, feel free to ask anything if needed.
>> >
>> > Cheers
>> >
>> > On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
>> >
>> >> Yeah, I changed the MTU to 9000 on the Guest network interface and
>> >> connectivity is now fine out of the box.
>> >> Hopefully I won't need to set jumbo frames in the VMs any time soon.
>> :-)
>> >>
>> >> Thanks for the tips.
>> >>
>> >> Lucian
>> >>
>> >> --
>> >> Sent from the Delta quadrant using Borg technology!
>> >>
>> >> Nux!
>> >> www.nux.ro
>> >>
>> >> ----- Original Message -----
>> >> > From: "Simon Weller" <sw...@ena.com.INVALID>
>> >> > To: "users" <us...@cloudstack.apache.org>
>> >> > Cc: "dev" <de...@cloudstack.apache.org>
>> >> > Sent: Monday, 20 November, 2017 17:36:31
>> >> > Subject: Re: Circumventing VXLAN MTU issues
>> >>
>> >> > Change your host interface MTU to something a lot higher.
>> >> >
>> >> >
>> >> > ________________________________
>> >> > From: Nux! <nu...@li.nux.ro>
>> >> > Sent: Monday, November 20, 2017 10:51 AM
>> >> > To: users
>> >> > Cc: dev
>> >> > Subject: Re: Circumventing VXLAN MTU issues
>> >> >
>> >> > Hi,
>> >> >
>> >> > I probably can use jumbo frames, but for now my lab is restricted to
>> a
>> >> single
>> >> > machine.
>> >> > Anything I can do in this situation?
>> >> >
>> >> > --
>> >> > Sent from the Delta quadrant using Borg technology!
>> >> >
>> >> > Nux!
>> >> > www.nux.ro
>> >> >
>> >> > ----- Original Message -----
>> >> >> From: "Simon Weller" <sw...@ena.com.INVALID>
>> >> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
>> >> users@cloudstack.apache.org>
>> >> >> Sent: Monday, 20 November, 2017 16:38:33
>> >> >> Subject: Re: Circumventing VXLAN MTU issues
>> >> >
>> >> >> Lucian,
>> >> >>
>> >> >> Can you run jumbos on your switches?
>> >> >>
>> >> >> - Si
>> >> >>
>> >> >>
>> >> >> ________________________________
>> >> >> From: Nux! <nu...@li.nux.ro>
>> >> >> Sent: Monday, November 20, 2017 10:08 AM
>> >> >> To: dev
>> >> >> Cc: users
>> >> >> Subject: Circumventing VXLAN MTU issues
>> >> >>
>> >> >> Hello,
>> >> >>
>> >> >> I am playing around with the native VXLAN implementation and I have
>> of
>> >> course
>> >> >> hit the situation where MTU on the host gets chomped by 50 bytes and
>> >> traffic
>> >> >> crawls to a stop in the VMs.
>> >> >> How are you people circumventing this? The easiest, but kind of
>> >> inconvenient for
>> >> >> customers would be to set MTU 1450 inside the VM, though it could be
>> >> baked into
>> >> >> the templates or via cloud-init.
>> >> >>
>> >> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
>> >> problem, I
>> >> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
>> >> the same
>> >> >> time this VXLAN implementation suffers from some limitations
>> (multicast
>> >> etc).
>> >> >>
>> >> >> Lucian
>> >> >>
>> >> >>
>> >> >> --
>> >> >> Sent from the Delta quadrant using Borg technology!
>> >> >>
>> >> >> Nux!
>> >> > > www.nux.ro
>> >>
>> >
>> >
>> >
>> > --
>> >
>> > Andrija Panić
>>
>
>
>
> --
>
> Andrija Panić
>
--
Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Andrija Panic <an...@gmail.com>.
@devs any plan to implement VXLAN as isolation for also Private Gateway
functionality ?
On 21 December 2017 at 15:28, Andrija Panic <an...@gmail.com> wrote:
> while there, here is another one (just below that one) :) which I learned
> the very hard way, after 2 years in production (when clients started
> complaining on random traffic drop) : http://docs.cloudstack.
> apache.org/en/latest/networking/vxlan.html#important-note-on-max-number-
> of-multicast-groups-and-thus-vxlan-intefaces
>
> Very sneaky and ugly: Ubuntu kernel 3.16 = random traffic drops, but all
> vxlan interfaces are UP etc..., Ubuntu kernel 4.x vxlan interface would NOT
> come up AT ALL, so a hard drop of packets, and there is where we saw the
> issue...
>
> Cheers
>
>
> On 21 December 2017 at 15:13, Nux! <nu...@li.nux.ro> wrote:
>
>> Thanks Andrija, well done, I have indeed read your document.
>> Using MTU 9000 solved my problem. :)
>>
>> Lucian
>>
>> --
>> Sent from the Delta quadrant using Borg technology!
>>
>> Nux!
>> www.nux.ro
>>
>> ----- Original Message -----
>> > From: "Andrija Panic" <an...@gmail.com>
>> > To: "dev" <de...@cloudstack.apache.org>
>> > Cc: "users" <us...@cloudstack.apache.org>
>> > Sent: Thursday, 21 December, 2017 14:09:43
>> > Subject: Re: Circumventing VXLAN MTU issues
>>
>> > Hi Nux,
>> >
>> > there is one rare contributions from my side :D to the ACS
>> (documentation,
>> > 2.5 years ago) - check it here:
>> > http://docs.cloudstack.apache.org/en/latest/networking/vxlan
>> .html#important-note-on-mtu-size
>> >
>> > We are using it extensively, feel free to ask anything if needed.
>> >
>> > Cheers
>> >
>> > On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
>> >
>> >> Yeah, I changed the MTU to 9000 on the Guest network interface and
>> >> connectivity is now fine out of the box.
>> >> Hopefully I won't need to set jumbo frames in the VMs any time soon.
>> :-)
>> >>
>> >> Thanks for the tips.
>> >>
>> >> Lucian
>> >>
>> >> --
>> >> Sent from the Delta quadrant using Borg technology!
>> >>
>> >> Nux!
>> >> www.nux.ro
>> >>
>> >> ----- Original Message -----
>> >> > From: "Simon Weller" <sw...@ena.com.INVALID>
>> >> > To: "users" <us...@cloudstack.apache.org>
>> >> > Cc: "dev" <de...@cloudstack.apache.org>
>> >> > Sent: Monday, 20 November, 2017 17:36:31
>> >> > Subject: Re: Circumventing VXLAN MTU issues
>> >>
>> >> > Change your host interface MTU to something a lot higher.
>> >> >
>> >> >
>> >> > ________________________________
>> >> > From: Nux! <nu...@li.nux.ro>
>> >> > Sent: Monday, November 20, 2017 10:51 AM
>> >> > To: users
>> >> > Cc: dev
>> >> > Subject: Re: Circumventing VXLAN MTU issues
>> >> >
>> >> > Hi,
>> >> >
>> >> > I probably can use jumbo frames, but for now my lab is restricted to
>> a
>> >> single
>> >> > machine.
>> >> > Anything I can do in this situation?
>> >> >
>> >> > --
>> >> > Sent from the Delta quadrant using Borg technology!
>> >> >
>> >> > Nux!
>> >> > www.nux.ro
>> >> >
>> >> > ----- Original Message -----
>> >> >> From: "Simon Weller" <sw...@ena.com.INVALID>
>> >> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
>> >> users@cloudstack.apache.org>
>> >> >> Sent: Monday, 20 November, 2017 16:38:33
>> >> >> Subject: Re: Circumventing VXLAN MTU issues
>> >> >
>> >> >> Lucian,
>> >> >>
>> >> >> Can you run jumbos on your switches?
>> >> >>
>> >> >> - Si
>> >> >>
>> >> >>
>> >> >> ________________________________
>> >> >> From: Nux! <nu...@li.nux.ro>
>> >> >> Sent: Monday, November 20, 2017 10:08 AM
>> >> >> To: dev
>> >> >> Cc: users
>> >> >> Subject: Circumventing VXLAN MTU issues
>> >> >>
>> >> >> Hello,
>> >> >>
>> >> >> I am playing around with the native VXLAN implementation and I have
>> of
>> >> course
>> >> >> hit the situation where MTU on the host gets chomped by 50 bytes and
>> >> traffic
>> >> >> crawls to a stop in the VMs.
>> >> >> How are you people circumventing this? The easiest, but kind of
>> >> inconvenient for
>> >> >> customers would be to set MTU 1450 inside the VM, though it could be
>> >> baked into
>> >> >> the templates or via cloud-init.
>> >> >>
>> >> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
>> >> problem, I
>> >> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
>> >> the same
>> >> >> time this VXLAN implementation suffers from some limitations
>> (multicast
>> >> etc).
>> >> >>
>> >> >> Lucian
>> >> >>
>> >> >>
>> >> >> --
>> >> >> Sent from the Delta quadrant using Borg technology!
>> >> >>
>> >> >> Nux!
>> >> > > www.nux.ro
>> >>
>> >
>> >
>> >
>> > --
>> >
>> > Andrija Panić
>>
>
>
>
> --
>
> Andrija Panić
>
--
Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Andrija Panic <an...@gmail.com>.
while there, here is another one (just below that one) :) which I learned
the very hard way, after 2 years in production (when clients started
complaining on random traffic drop) :
http://docs.cloudstack.apache.org/en/latest/networking/vxlan.html#important-note-on-max-number-of-multicast-groups-and-thus-vxlan-intefaces
Very sneaky and ugly: Ubuntu kernel 3.16 = random traffic drops, but all
vxlan interfaces are UP etc..., Ubuntu kernel 4.x vxlan interface would NOT
come up AT ALL, so a hard drop of packets, and there is where we saw the
issue...
Cheers
On 21 December 2017 at 15:13, Nux! <nu...@li.nux.ro> wrote:
> Thanks Andrija, well done, I have indeed read your document.
> Using MTU 9000 solved my problem. :)
>
> Lucian
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
>
> ----- Original Message -----
> > From: "Andrija Panic" <an...@gmail.com>
> > To: "dev" <de...@cloudstack.apache.org>
> > Cc: "users" <us...@cloudstack.apache.org>
> > Sent: Thursday, 21 December, 2017 14:09:43
> > Subject: Re: Circumventing VXLAN MTU issues
>
> > Hi Nux,
> >
> > there is one rare contributions from my side :D to the ACS
> (documentation,
> > 2.5 years ago) - check it here:
> > http://docs.cloudstack.apache.org/en/latest/networking/
> vxlan.html#important-note-on-mtu-size
> >
> > We are using it extensively, feel free to ask anything if needed.
> >
> > Cheers
> >
> > On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
> >
> >> Yeah, I changed the MTU to 9000 on the Guest network interface and
> >> connectivity is now fine out of the box.
> >> Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
> >>
> >> Thanks for the tips.
> >>
> >> Lucian
> >>
> >> --
> >> Sent from the Delta quadrant using Borg technology!
> >>
> >> Nux!
> >> www.nux.ro
> >>
> >> ----- Original Message -----
> >> > From: "Simon Weller" <sw...@ena.com.INVALID>
> >> > To: "users" <us...@cloudstack.apache.org>
> >> > Cc: "dev" <de...@cloudstack.apache.org>
> >> > Sent: Monday, 20 November, 2017 17:36:31
> >> > Subject: Re: Circumventing VXLAN MTU issues
> >>
> >> > Change your host interface MTU to something a lot higher.
> >> >
> >> >
> >> > ________________________________
> >> > From: Nux! <nu...@li.nux.ro>
> >> > Sent: Monday, November 20, 2017 10:51 AM
> >> > To: users
> >> > Cc: dev
> >> > Subject: Re: Circumventing VXLAN MTU issues
> >> >
> >> > Hi,
> >> >
> >> > I probably can use jumbo frames, but for now my lab is restricted to a
> >> single
> >> > machine.
> >> > Anything I can do in this situation?
> >> >
> >> > --
> >> > Sent from the Delta quadrant using Borg technology!
> >> >
> >> > Nux!
> >> > www.nux.ro
> >> >
> >> > ----- Original Message -----
> >> >> From: "Simon Weller" <sw...@ena.com.INVALID>
> >> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
> >> users@cloudstack.apache.org>
> >> >> Sent: Monday, 20 November, 2017 16:38:33
> >> >> Subject: Re: Circumventing VXLAN MTU issues
> >> >
> >> >> Lucian,
> >> >>
> >> >> Can you run jumbos on your switches?
> >> >>
> >> >> - Si
> >> >>
> >> >>
> >> >> ________________________________
> >> >> From: Nux! <nu...@li.nux.ro>
> >> >> Sent: Monday, November 20, 2017 10:08 AM
> >> >> To: dev
> >> >> Cc: users
> >> >> Subject: Circumventing VXLAN MTU issues
> >> >>
> >> >> Hello,
> >> >>
> >> >> I am playing around with the native VXLAN implementation and I have
> of
> >> course
> >> >> hit the situation where MTU on the host gets chomped by 50 bytes and
> >> traffic
> >> >> crawls to a stop in the VMs.
> >> >> How are you people circumventing this? The easiest, but kind of
> >> inconvenient for
> >> >> customers would be to set MTU 1450 inside the VM, though it could be
> >> baked into
> >> >> the templates or via cloud-init.
> >> >>
> >> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
> >> problem, I
> >> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
> >> the same
> >> >> time this VXLAN implementation suffers from some limitations
> (multicast
> >> etc).
> >> >>
> >> >> Lucian
> >> >>
> >> >>
> >> >> --
> >> >> Sent from the Delta quadrant using Borg technology!
> >> >>
> >> >> Nux!
> >> > > www.nux.ro
> >>
> >
> >
> >
> > --
> >
> > Andrija Panić
>
--
Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Andrija Panic <an...@gmail.com>.
while there, here is another one (just below that one) :) which I learned
the very hard way, after 2 years in production (when clients started
complaining on random traffic drop) :
http://docs.cloudstack.apache.org/en/latest/networking/vxlan.html#important-note-on-max-number-of-multicast-groups-and-thus-vxlan-intefaces
Very sneaky and ugly: Ubuntu kernel 3.16 = random traffic drops, but all
vxlan interfaces are UP etc..., Ubuntu kernel 4.x vxlan interface would NOT
come up AT ALL, so a hard drop of packets, and there is where we saw the
issue...
Cheers
On 21 December 2017 at 15:13, Nux! <nu...@li.nux.ro> wrote:
> Thanks Andrija, well done, I have indeed read your document.
> Using MTU 9000 solved my problem. :)
>
> Lucian
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
>
> ----- Original Message -----
> > From: "Andrija Panic" <an...@gmail.com>
> > To: "dev" <de...@cloudstack.apache.org>
> > Cc: "users" <us...@cloudstack.apache.org>
> > Sent: Thursday, 21 December, 2017 14:09:43
> > Subject: Re: Circumventing VXLAN MTU issues
>
> > Hi Nux,
> >
> > there is one rare contributions from my side :D to the ACS
> (documentation,
> > 2.5 years ago) - check it here:
> > http://docs.cloudstack.apache.org/en/latest/networking/
> vxlan.html#important-note-on-mtu-size
> >
> > We are using it extensively, feel free to ask anything if needed.
> >
> > Cheers
> >
> > On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
> >
> >> Yeah, I changed the MTU to 9000 on the Guest network interface and
> >> connectivity is now fine out of the box.
> >> Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
> >>
> >> Thanks for the tips.
> >>
> >> Lucian
> >>
> >> --
> >> Sent from the Delta quadrant using Borg technology!
> >>
> >> Nux!
> >> www.nux.ro
> >>
> >> ----- Original Message -----
> >> > From: "Simon Weller" <sw...@ena.com.INVALID>
> >> > To: "users" <us...@cloudstack.apache.org>
> >> > Cc: "dev" <de...@cloudstack.apache.org>
> >> > Sent: Monday, 20 November, 2017 17:36:31
> >> > Subject: Re: Circumventing VXLAN MTU issues
> >>
> >> > Change your host interface MTU to something a lot higher.
> >> >
> >> >
> >> > ________________________________
> >> > From: Nux! <nu...@li.nux.ro>
> >> > Sent: Monday, November 20, 2017 10:51 AM
> >> > To: users
> >> > Cc: dev
> >> > Subject: Re: Circumventing VXLAN MTU issues
> >> >
> >> > Hi,
> >> >
> >> > I probably can use jumbo frames, but for now my lab is restricted to a
> >> single
> >> > machine.
> >> > Anything I can do in this situation?
> >> >
> >> > --
> >> > Sent from the Delta quadrant using Borg technology!
> >> >
> >> > Nux!
> >> > www.nux.ro
> >> >
> >> > ----- Original Message -----
> >> >> From: "Simon Weller" <sw...@ena.com.INVALID>
> >> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
> >> users@cloudstack.apache.org>
> >> >> Sent: Monday, 20 November, 2017 16:38:33
> >> >> Subject: Re: Circumventing VXLAN MTU issues
> >> >
> >> >> Lucian,
> >> >>
> >> >> Can you run jumbos on your switches?
> >> >>
> >> >> - Si
> >> >>
> >> >>
> >> >> ________________________________
> >> >> From: Nux! <nu...@li.nux.ro>
> >> >> Sent: Monday, November 20, 2017 10:08 AM
> >> >> To: dev
> >> >> Cc: users
> >> >> Subject: Circumventing VXLAN MTU issues
> >> >>
> >> >> Hello,
> >> >>
> >> >> I am playing around with the native VXLAN implementation and I have
> of
> >> course
> >> >> hit the situation where MTU on the host gets chomped by 50 bytes and
> >> traffic
> >> >> crawls to a stop in the VMs.
> >> >> How are you people circumventing this? The easiest, but kind of
> >> inconvenient for
> >> >> customers would be to set MTU 1450 inside the VM, though it could be
> >> baked into
> >> >> the templates or via cloud-init.
> >> >>
> >> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
> >> problem, I
> >> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
> >> the same
> >> >> time this VXLAN implementation suffers from some limitations
> (multicast
> >> etc).
> >> >>
> >> >> Lucian
> >> >>
> >> >>
> >> >> --
> >> >> Sent from the Delta quadrant using Borg technology!
> >> >>
> >> >> Nux!
> >> > > www.nux.ro
> >>
> >
> >
> >
> > --
> >
> > Andrija Panić
>
--
Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Nux! <nu...@li.nux.ro>.
Thanks Andrija, well done, I have indeed read your document.
Using MTU 9000 solved my problem. :)
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Andrija Panic" <an...@gmail.com>
> To: "dev" <de...@cloudstack.apache.org>
> Cc: "users" <us...@cloudstack.apache.org>
> Sent: Thursday, 21 December, 2017 14:09:43
> Subject: Re: Circumventing VXLAN MTU issues
> Hi Nux,
>
> there is one rare contributions from my side :D to the ACS (documentation,
> 2.5 years ago) - check it here:
> http://docs.cloudstack.apache.org/en/latest/networking/vxlan.html#important-note-on-mtu-size
>
> We are using it extensively, feel free to ask anything if needed.
>
> Cheers
>
> On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
>
>> Yeah, I changed the MTU to 9000 on the Guest network interface and
>> connectivity is now fine out of the box.
>> Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
>>
>> Thanks for the tips.
>>
>> Lucian
>>
>> --
>> Sent from the Delta quadrant using Borg technology!
>>
>> Nux!
>> www.nux.ro
>>
>> ----- Original Message -----
>> > From: "Simon Weller" <sw...@ena.com.INVALID>
>> > To: "users" <us...@cloudstack.apache.org>
>> > Cc: "dev" <de...@cloudstack.apache.org>
>> > Sent: Monday, 20 November, 2017 17:36:31
>> > Subject: Re: Circumventing VXLAN MTU issues
>>
>> > Change your host interface MTU to something a lot higher.
>> >
>> >
>> > ________________________________
>> > From: Nux! <nu...@li.nux.ro>
>> > Sent: Monday, November 20, 2017 10:51 AM
>> > To: users
>> > Cc: dev
>> > Subject: Re: Circumventing VXLAN MTU issues
>> >
>> > Hi,
>> >
>> > I probably can use jumbo frames, but for now my lab is restricted to a
>> single
>> > machine.
>> > Anything I can do in this situation?
>> >
>> > --
>> > Sent from the Delta quadrant using Borg technology!
>> >
>> > Nux!
>> > www.nux.ro
>> >
>> > ----- Original Message -----
>> >> From: "Simon Weller" <sw...@ena.com.INVALID>
>> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
>> users@cloudstack.apache.org>
>> >> Sent: Monday, 20 November, 2017 16:38:33
>> >> Subject: Re: Circumventing VXLAN MTU issues
>> >
>> >> Lucian,
>> >>
>> >> Can you run jumbos on your switches?
>> >>
>> >> - Si
>> >>
>> >>
>> >> ________________________________
>> >> From: Nux! <nu...@li.nux.ro>
>> >> Sent: Monday, November 20, 2017 10:08 AM
>> >> To: dev
>> >> Cc: users
>> >> Subject: Circumventing VXLAN MTU issues
>> >>
>> >> Hello,
>> >>
>> >> I am playing around with the native VXLAN implementation and I have of
>> course
>> >> hit the situation where MTU on the host gets chomped by 50 bytes and
>> traffic
>> >> crawls to a stop in the VMs.
>> >> How are you people circumventing this? The easiest, but kind of
>> inconvenient for
>> >> customers would be to set MTU 1450 inside the VM, though it could be
>> baked into
>> >> the templates or via cloud-init.
>> >>
>> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
>> problem, I
>> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
>> the same
>> >> time this VXLAN implementation suffers from some limitations (multicast
>> etc).
>> >>
>> >> Lucian
>> >>
>> >>
>> >> --
>> >> Sent from the Delta quadrant using Borg technology!
>> >>
>> >> Nux!
>> > > www.nux.ro
>>
>
>
>
> --
>
> Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Nux! <nu...@li.nux.ro>.
Thanks Andrija, well done, I have indeed read your document.
Using MTU 9000 solved my problem. :)
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Andrija Panic" <an...@gmail.com>
> To: "dev" <de...@cloudstack.apache.org>
> Cc: "users" <us...@cloudstack.apache.org>
> Sent: Thursday, 21 December, 2017 14:09:43
> Subject: Re: Circumventing VXLAN MTU issues
> Hi Nux,
>
> there is one rare contributions from my side :D to the ACS (documentation,
> 2.5 years ago) - check it here:
> http://docs.cloudstack.apache.org/en/latest/networking/vxlan.html#important-note-on-mtu-size
>
> We are using it extensively, feel free to ask anything if needed.
>
> Cheers
>
> On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
>
>> Yeah, I changed the MTU to 9000 on the Guest network interface and
>> connectivity is now fine out of the box.
>> Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
>>
>> Thanks for the tips.
>>
>> Lucian
>>
>> --
>> Sent from the Delta quadrant using Borg technology!
>>
>> Nux!
>> www.nux.ro
>>
>> ----- Original Message -----
>> > From: "Simon Weller" <sw...@ena.com.INVALID>
>> > To: "users" <us...@cloudstack.apache.org>
>> > Cc: "dev" <de...@cloudstack.apache.org>
>> > Sent: Monday, 20 November, 2017 17:36:31
>> > Subject: Re: Circumventing VXLAN MTU issues
>>
>> > Change your host interface MTU to something a lot higher.
>> >
>> >
>> > ________________________________
>> > From: Nux! <nu...@li.nux.ro>
>> > Sent: Monday, November 20, 2017 10:51 AM
>> > To: users
>> > Cc: dev
>> > Subject: Re: Circumventing VXLAN MTU issues
>> >
>> > Hi,
>> >
>> > I probably can use jumbo frames, but for now my lab is restricted to a
>> single
>> > machine.
>> > Anything I can do in this situation?
>> >
>> > --
>> > Sent from the Delta quadrant using Borg technology!
>> >
>> > Nux!
>> > www.nux.ro
>> >
>> > ----- Original Message -----
>> >> From: "Simon Weller" <sw...@ena.com.INVALID>
>> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
>> users@cloudstack.apache.org>
>> >> Sent: Monday, 20 November, 2017 16:38:33
>> >> Subject: Re: Circumventing VXLAN MTU issues
>> >
>> >> Lucian,
>> >>
>> >> Can you run jumbos on your switches?
>> >>
>> >> - Si
>> >>
>> >>
>> >> ________________________________
>> >> From: Nux! <nu...@li.nux.ro>
>> >> Sent: Monday, November 20, 2017 10:08 AM
>> >> To: dev
>> >> Cc: users
>> >> Subject: Circumventing VXLAN MTU issues
>> >>
>> >> Hello,
>> >>
>> >> I am playing around with the native VXLAN implementation and I have of
>> course
>> >> hit the situation where MTU on the host gets chomped by 50 bytes and
>> traffic
>> >> crawls to a stop in the VMs.
>> >> How are you people circumventing this? The easiest, but kind of
>> inconvenient for
>> >> customers would be to set MTU 1450 inside the VM, though it could be
>> baked into
>> >> the templates or via cloud-init.
>> >>
>> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
>> problem, I
>> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
>> the same
>> >> time this VXLAN implementation suffers from some limitations (multicast
>> etc).
>> >>
>> >> Lucian
>> >>
>> >>
>> >> --
>> >> Sent from the Delta quadrant using Borg technology!
>> >>
>> >> Nux!
>> > > www.nux.ro
>>
>
>
>
> --
>
> Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Andrija Panic <an...@gmail.com>.
Hi Nux,
there is one rare contributions from my side :D to the ACS (documentation,
2.5 years ago) - check it here:
http://docs.cloudstack.apache.org/en/latest/networking/vxlan.html#important-note-on-mtu-size
We are using it extensively, feel free to ask anything if needed.
Cheers
On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
> Yeah, I changed the MTU to 9000 on the Guest network interface and
> connectivity is now fine out of the box.
> Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
>
> Thanks for the tips.
>
> Lucian
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
>
> ----- Original Message -----
> > From: "Simon Weller" <sw...@ena.com.INVALID>
> > To: "users" <us...@cloudstack.apache.org>
> > Cc: "dev" <de...@cloudstack.apache.org>
> > Sent: Monday, 20 November, 2017 17:36:31
> > Subject: Re: Circumventing VXLAN MTU issues
>
> > Change your host interface MTU to something a lot higher.
> >
> >
> > ________________________________
> > From: Nux! <nu...@li.nux.ro>
> > Sent: Monday, November 20, 2017 10:51 AM
> > To: users
> > Cc: dev
> > Subject: Re: Circumventing VXLAN MTU issues
> >
> > Hi,
> >
> > I probably can use jumbo frames, but for now my lab is restricted to a
> single
> > machine.
> > Anything I can do in this situation?
> >
> > --
> > Sent from the Delta quadrant using Borg technology!
> >
> > Nux!
> > www.nux.ro
> >
> > ----- Original Message -----
> >> From: "Simon Weller" <sw...@ena.com.INVALID>
> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
> users@cloudstack.apache.org>
> >> Sent: Monday, 20 November, 2017 16:38:33
> >> Subject: Re: Circumventing VXLAN MTU issues
> >
> >> Lucian,
> >>
> >> Can you run jumbos on your switches?
> >>
> >> - Si
> >>
> >>
> >> ________________________________
> >> From: Nux! <nu...@li.nux.ro>
> >> Sent: Monday, November 20, 2017 10:08 AM
> >> To: dev
> >> Cc: users
> >> Subject: Circumventing VXLAN MTU issues
> >>
> >> Hello,
> >>
> >> I am playing around with the native VXLAN implementation and I have of
> course
> >> hit the situation where MTU on the host gets chomped by 50 bytes and
> traffic
> >> crawls to a stop in the VMs.
> >> How are you people circumventing this? The easiest, but kind of
> inconvenient for
> >> customers would be to set MTU 1450 inside the VM, though it could be
> baked into
> >> the templates or via cloud-init.
> >>
> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
> problem, I
> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
> the same
> >> time this VXLAN implementation suffers from some limitations (multicast
> etc).
> >>
> >> Lucian
> >>
> >>
> >> --
> >> Sent from the Delta quadrant using Borg technology!
> >>
> >> Nux!
> > > www.nux.ro
>
--
Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Andrija Panic <an...@gmail.com>.
Hi Nux,
there is one rare contributions from my side :D to the ACS (documentation,
2.5 years ago) - check it here:
http://docs.cloudstack.apache.org/en/latest/networking/vxlan.html#important-note-on-mtu-size
We are using it extensively, feel free to ask anything if needed.
Cheers
On 20 November 2017 at 19:38, Nux! <nu...@li.nux.ro> wrote:
> Yeah, I changed the MTU to 9000 on the Guest network interface and
> connectivity is now fine out of the box.
> Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
>
> Thanks for the tips.
>
> Lucian
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
>
> ----- Original Message -----
> > From: "Simon Weller" <sw...@ena.com.INVALID>
> > To: "users" <us...@cloudstack.apache.org>
> > Cc: "dev" <de...@cloudstack.apache.org>
> > Sent: Monday, 20 November, 2017 17:36:31
> > Subject: Re: Circumventing VXLAN MTU issues
>
> > Change your host interface MTU to something a lot higher.
> >
> >
> > ________________________________
> > From: Nux! <nu...@li.nux.ro>
> > Sent: Monday, November 20, 2017 10:51 AM
> > To: users
> > Cc: dev
> > Subject: Re: Circumventing VXLAN MTU issues
> >
> > Hi,
> >
> > I probably can use jumbo frames, but for now my lab is restricted to a
> single
> > machine.
> > Anything I can do in this situation?
> >
> > --
> > Sent from the Delta quadrant using Borg technology!
> >
> > Nux!
> > www.nux.ro
> >
> > ----- Original Message -----
> >> From: "Simon Weller" <sw...@ena.com.INVALID>
> >> To: "dev" <de...@cloudstack.apache.org>, "users" <
> users@cloudstack.apache.org>
> >> Sent: Monday, 20 November, 2017 16:38:33
> >> Subject: Re: Circumventing VXLAN MTU issues
> >
> >> Lucian,
> >>
> >> Can you run jumbos on your switches?
> >>
> >> - Si
> >>
> >>
> >> ________________________________
> >> From: Nux! <nu...@li.nux.ro>
> >> Sent: Monday, November 20, 2017 10:08 AM
> >> To: dev
> >> Cc: users
> >> Subject: Circumventing VXLAN MTU issues
> >>
> >> Hello,
> >>
> >> I am playing around with the native VXLAN implementation and I have of
> course
> >> hit the situation where MTU on the host gets chomped by 50 bytes and
> traffic
> >> crawls to a stop in the VMs.
> >> How are you people circumventing this? The easiest, but kind of
> inconvenient for
> >> customers would be to set MTU 1450 inside the VM, though it could be
> baked into
> >> the templates or via cloud-init.
> >>
> >> Is the OpenVSwitch implementation (of VXLAN) suffer from the same
> problem, I
> >> understand ovs can dynamically adjust MTUs to compensate (?), but at
> the same
> >> time this VXLAN implementation suffers from some limitations (multicast
> etc).
> >>
> >> Lucian
> >>
> >>
> >> --
> >> Sent from the Delta quadrant using Borg technology!
> >>
> >> Nux!
> > > www.nux.ro
>
--
Andrija Panić
Re: Circumventing VXLAN MTU issues
Posted by Nux! <nu...@li.nux.ro>.
Yeah, I changed the MTU to 9000 on the Guest network interface and connectivity is now fine out of the box.
Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
Thanks for the tips.
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Simon Weller" <sw...@ena.com.INVALID>
> To: "users" <us...@cloudstack.apache.org>
> Cc: "dev" <de...@cloudstack.apache.org>
> Sent: Monday, 20 November, 2017 17:36:31
> Subject: Re: Circumventing VXLAN MTU issues
> Change your host interface MTU to something a lot higher.
>
>
> ________________________________
> From: Nux! <nu...@li.nux.ro>
> Sent: Monday, November 20, 2017 10:51 AM
> To: users
> Cc: dev
> Subject: Re: Circumventing VXLAN MTU issues
>
> Hi,
>
> I probably can use jumbo frames, but for now my lab is restricted to a single
> machine.
> Anything I can do in this situation?
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
>
> ----- Original Message -----
>> From: "Simon Weller" <sw...@ena.com.INVALID>
>> To: "dev" <de...@cloudstack.apache.org>, "users" <us...@cloudstack.apache.org>
>> Sent: Monday, 20 November, 2017 16:38:33
>> Subject: Re: Circumventing VXLAN MTU issues
>
>> Lucian,
>>
>> Can you run jumbos on your switches?
>>
>> - Si
>>
>>
>> ________________________________
>> From: Nux! <nu...@li.nux.ro>
>> Sent: Monday, November 20, 2017 10:08 AM
>> To: dev
>> Cc: users
>> Subject: Circumventing VXLAN MTU issues
>>
>> Hello,
>>
>> I am playing around with the native VXLAN implementation and I have of course
>> hit the situation where MTU on the host gets chomped by 50 bytes and traffic
>> crawls to a stop in the VMs.
>> How are you people circumventing this? The easiest, but kind of inconvenient for
>> customers would be to set MTU 1450 inside the VM, though it could be baked into
>> the templates or via cloud-init.
>>
>> Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I
>> understand ovs can dynamically adjust MTUs to compensate (?), but at the same
>> time this VXLAN implementation suffers from some limitations (multicast etc).
>>
>> Lucian
>>
>>
>> --
>> Sent from the Delta quadrant using Borg technology!
>>
>> Nux!
> > www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Nux! <nu...@li.nux.ro>.
Yeah, I changed the MTU to 9000 on the Guest network interface and connectivity is now fine out of the box.
Hopefully I won't need to set jumbo frames in the VMs any time soon. :-)
Thanks for the tips.
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Simon Weller" <sw...@ena.com.INVALID>
> To: "users" <us...@cloudstack.apache.org>
> Cc: "dev" <de...@cloudstack.apache.org>
> Sent: Monday, 20 November, 2017 17:36:31
> Subject: Re: Circumventing VXLAN MTU issues
> Change your host interface MTU to something a lot higher.
>
>
> ________________________________
> From: Nux! <nu...@li.nux.ro>
> Sent: Monday, November 20, 2017 10:51 AM
> To: users
> Cc: dev
> Subject: Re: Circumventing VXLAN MTU issues
>
> Hi,
>
> I probably can use jumbo frames, but for now my lab is restricted to a single
> machine.
> Anything I can do in this situation?
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
>
> ----- Original Message -----
>> From: "Simon Weller" <sw...@ena.com.INVALID>
>> To: "dev" <de...@cloudstack.apache.org>, "users" <us...@cloudstack.apache.org>
>> Sent: Monday, 20 November, 2017 16:38:33
>> Subject: Re: Circumventing VXLAN MTU issues
>
>> Lucian,
>>
>> Can you run jumbos on your switches?
>>
>> - Si
>>
>>
>> ________________________________
>> From: Nux! <nu...@li.nux.ro>
>> Sent: Monday, November 20, 2017 10:08 AM
>> To: dev
>> Cc: users
>> Subject: Circumventing VXLAN MTU issues
>>
>> Hello,
>>
>> I am playing around with the native VXLAN implementation and I have of course
>> hit the situation where MTU on the host gets chomped by 50 bytes and traffic
>> crawls to a stop in the VMs.
>> How are you people circumventing this? The easiest, but kind of inconvenient for
>> customers would be to set MTU 1450 inside the VM, though it could be baked into
>> the templates or via cloud-init.
>>
>> Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I
>> understand ovs can dynamically adjust MTUs to compensate (?), but at the same
>> time this VXLAN implementation suffers from some limitations (multicast etc).
>>
>> Lucian
>>
>>
>> --
>> Sent from the Delta quadrant using Borg technology!
>>
>> Nux!
> > www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Simon Weller <sw...@ena.com.INVALID>.
Change your host interface MTU to something a lot higher.
________________________________
From: Nux! <nu...@li.nux.ro>
Sent: Monday, November 20, 2017 10:51 AM
To: users
Cc: dev
Subject: Re: Circumventing VXLAN MTU issues
Hi,
I probably can use jumbo frames, but for now my lab is restricted to a single machine.
Anything I can do in this situation?
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Simon Weller" <sw...@ena.com.INVALID>
> To: "dev" <de...@cloudstack.apache.org>, "users" <us...@cloudstack.apache.org>
> Sent: Monday, 20 November, 2017 16:38:33
> Subject: Re: Circumventing VXLAN MTU issues
> Lucian,
>
> Can you run jumbos on your switches?
>
> - Si
>
>
> ________________________________
> From: Nux! <nu...@li.nux.ro>
> Sent: Monday, November 20, 2017 10:08 AM
> To: dev
> Cc: users
> Subject: Circumventing VXLAN MTU issues
>
> Hello,
>
> I am playing around with the native VXLAN implementation and I have of course
> hit the situation where MTU on the host gets chomped by 50 bytes and traffic
> crawls to a stop in the VMs.
> How are you people circumventing this? The easiest, but kind of inconvenient for
> customers would be to set MTU 1450 inside the VM, though it could be baked into
> the templates or via cloud-init.
>
> Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I
> understand ovs can dynamically adjust MTUs to compensate (?), but at the same
> time this VXLAN implementation suffers from some limitations (multicast etc).
>
> Lucian
>
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Simon Weller <sw...@ena.com.INVALID>.
Change your host interface MTU to something a lot higher.
________________________________
From: Nux! <nu...@li.nux.ro>
Sent: Monday, November 20, 2017 10:51 AM
To: users
Cc: dev
Subject: Re: Circumventing VXLAN MTU issues
Hi,
I probably can use jumbo frames, but for now my lab is restricted to a single machine.
Anything I can do in this situation?
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Simon Weller" <sw...@ena.com.INVALID>
> To: "dev" <de...@cloudstack.apache.org>, "users" <us...@cloudstack.apache.org>
> Sent: Monday, 20 November, 2017 16:38:33
> Subject: Re: Circumventing VXLAN MTU issues
> Lucian,
>
> Can you run jumbos on your switches?
>
> - Si
>
>
> ________________________________
> From: Nux! <nu...@li.nux.ro>
> Sent: Monday, November 20, 2017 10:08 AM
> To: dev
> Cc: users
> Subject: Circumventing VXLAN MTU issues
>
> Hello,
>
> I am playing around with the native VXLAN implementation and I have of course
> hit the situation where MTU on the host gets chomped by 50 bytes and traffic
> crawls to a stop in the VMs.
> How are you people circumventing this? The easiest, but kind of inconvenient for
> customers would be to set MTU 1450 inside the VM, though it could be baked into
> the templates or via cloud-init.
>
> Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I
> understand ovs can dynamically adjust MTUs to compensate (?), but at the same
> time this VXLAN implementation suffers from some limitations (multicast etc).
>
> Lucian
>
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Nux! <nu...@li.nux.ro>.
Hi,
I probably can use jumbo frames, but for now my lab is restricted to a single machine.
Anything I can do in this situation?
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Simon Weller" <sw...@ena.com.INVALID>
> To: "dev" <de...@cloudstack.apache.org>, "users" <us...@cloudstack.apache.org>
> Sent: Monday, 20 November, 2017 16:38:33
> Subject: Re: Circumventing VXLAN MTU issues
> Lucian,
>
> Can you run jumbos on your switches?
>
> - Si
>
>
> ________________________________
> From: Nux! <nu...@li.nux.ro>
> Sent: Monday, November 20, 2017 10:08 AM
> To: dev
> Cc: users
> Subject: Circumventing VXLAN MTU issues
>
> Hello,
>
> I am playing around with the native VXLAN implementation and I have of course
> hit the situation where MTU on the host gets chomped by 50 bytes and traffic
> crawls to a stop in the VMs.
> How are you people circumventing this? The easiest, but kind of inconvenient for
> customers would be to set MTU 1450 inside the VM, though it could be baked into
> the templates or via cloud-init.
>
> Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I
> understand ovs can dynamically adjust MTUs to compensate (?), but at the same
> time this VXLAN implementation suffers from some limitations (multicast etc).
>
> Lucian
>
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Nux! <nu...@li.nux.ro>.
Hi,
I probably can use jumbo frames, but for now my lab is restricted to a single machine.
Anything I can do in this situation?
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
----- Original Message -----
> From: "Simon Weller" <sw...@ena.com.INVALID>
> To: "dev" <de...@cloudstack.apache.org>, "users" <us...@cloudstack.apache.org>
> Sent: Monday, 20 November, 2017 16:38:33
> Subject: Re: Circumventing VXLAN MTU issues
> Lucian,
>
> Can you run jumbos on your switches?
>
> - Si
>
>
> ________________________________
> From: Nux! <nu...@li.nux.ro>
> Sent: Monday, November 20, 2017 10:08 AM
> To: dev
> Cc: users
> Subject: Circumventing VXLAN MTU issues
>
> Hello,
>
> I am playing around with the native VXLAN implementation and I have of course
> hit the situation where MTU on the host gets chomped by 50 bytes and traffic
> crawls to a stop in the VMs.
> How are you people circumventing this? The easiest, but kind of inconvenient for
> customers would be to set MTU 1450 inside the VM, though it could be baked into
> the templates or via cloud-init.
>
> Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I
> understand ovs can dynamically adjust MTUs to compensate (?), but at the same
> time this VXLAN implementation suffers from some limitations (multicast etc).
>
> Lucian
>
>
> --
> Sent from the Delta quadrant using Borg technology!
>
> Nux!
> www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Simon Weller <sw...@ena.com.INVALID>.
Lucian,
Can you run jumbos on your switches?
- Si
________________________________
From: Nux! <nu...@li.nux.ro>
Sent: Monday, November 20, 2017 10:08 AM
To: dev
Cc: users
Subject: Circumventing VXLAN MTU issues
Hello,
I am playing around with the native VXLAN implementation and I have of course hit the situation where MTU on the host gets chomped by 50 bytes and traffic crawls to a stop in the VMs.
How are you people circumventing this? The easiest, but kind of inconvenient for customers would be to set MTU 1450 inside the VM, though it could be baked into the templates or via cloud-init.
Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I understand ovs can dynamically adjust MTUs to compensate (?), but at the same time this VXLAN implementation suffers from some limitations (multicast etc).
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro
Re: Circumventing VXLAN MTU issues
Posted by Simon Weller <sw...@ena.com.INVALID>.
Lucian,
Can you run jumbos on your switches?
- Si
________________________________
From: Nux! <nu...@li.nux.ro>
Sent: Monday, November 20, 2017 10:08 AM
To: dev
Cc: users
Subject: Circumventing VXLAN MTU issues
Hello,
I am playing around with the native VXLAN implementation and I have of course hit the situation where MTU on the host gets chomped by 50 bytes and traffic crawls to a stop in the VMs.
How are you people circumventing this? The easiest, but kind of inconvenient for customers would be to set MTU 1450 inside the VM, though it could be baked into the templates or via cloud-init.
Is the OpenVSwitch implementation (of VXLAN) suffer from the same problem, I understand ovs can dynamically adjust MTUs to compensate (?), but at the same time this VXLAN implementation suffers from some limitations (multicast etc).
Lucian
--
Sent from the Delta quadrant using Borg technology!
Nux!
www.nux.ro