GRE and VXLAN.


Ghazisaidi, Navid <navid.ghazisaidi@...>
 

Hi,

I configured a very simple scenario, where two VMs on a server are connected to each other via veth. Each VM is interfacing with one bridge and bridges are controlled by controller. Overall, it is something like this: VM1—br1 (OVS1)—br2 (OVS2)—VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1 options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2 options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end, it mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I restarted controller and every time it learns network differently, once with 2 VMs connected to one OVS, another time with recognizing a wrong number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping each other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same observations.
Note 2. I tested the same topology with VXLAN and had the same observations.

I think these are the bugs and my question is GRE and VXLAN are supposed to be supported at all or not? If yes, is there any plan to fix these bugs? if not, is there any plan to support them?

Thanks,
Navid 


Luis Gomez
 

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have not tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release distributions had some performance issues.

https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Controller_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4 Cores

Let me know if this helps

BR/Luis

On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid <navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are connected to each other via veth. Each VM is interfacing with one bridge and bridges are controlled by controller. Overall, it is something like this: VM1—br1 (OVS1)—br2 (OVS2)—VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1 options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2 options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end, it mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I restarted controller and every time it learns network differently, once with 2 VMs connected to one OVS, another time with recognizing a wrong number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping each other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same observations.
Note 2. I tested the same topology with VXLAN and had the same observations.

I think these are the bugs and my question is GRE and VXLAN are supposed to be supported at all or not? If yes, is there any plan to fix these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Ghazisaidi, Navid <navid.ghazisaidi@...>
 

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores were
in used by controller.

Regards,
Navid

On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.

https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Controll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one bridge
and bridges are controlled by controller. Overall, it is something like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1 options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2 options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end, it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently, once
with 2 VMs connected to one OVS, another time with recognizing a wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Luis Gomez
 

SP edition with latest code, you will need -Xmx2g option when starting the controller as well.

BR/Luis

On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid <navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.

https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Controll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one bridge
and bridges are controlled by controller. Overall, it is something like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1 options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2 options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end, it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently, once
with 2 VMs connected to one OVS, another time with recognizing a wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Ghazisaidi, Navid <navid.ghazisaidi@...>
 

I just ran the latest one (0.1.2) and the GUI is not coming up at all.

Navid

On 5/1/14, 10:12 PM, "Luis Gomez" <ecelgp@...> wrote:

SP edition with latest code, you will need -Xmx2g option when starting
the controller as well.

BR/Luis


On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores
were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.


https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Contro
ll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one
bridge
and bridges are controlled by controller. Overall, it is something
like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1
options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2
options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end, it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently,
once
with 2 VMs connected to one OVS, another time with recognizing a wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping
each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to
fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Luis Gomez
 

Can you please send the OSGi console printout?

On May 2, 2014, at 10:25 AM, Ghazisaidi, Navid <navid.ghazisaidi@...> wrote:

I just ran the latest one (0.1.2) and the GUI is not coming up at all.

Navid



On 5/1/14, 10:12 PM, "Luis Gomez" <ecelgp@...> wrote:

SP edition with latest code, you will need -Xmx2g option when starting
the controller as well.

BR/Luis


On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores
were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.


https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Contro
ll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one
bridge
and bridges are controlled by controller. Overall, it is something
like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1
options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2
options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end, it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently,
once
with 2 VMs connected to one OVS, another time with recognizing a wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping
each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to
fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Ghazisaidi, Navid <navid.ghazisaidi@...>
 

Finally, I was able to bring up the GUI. It is not very stable. Sometimes
come up sometimes not….

Btw do you know who then knows if GRE and VXLAN are supposed to be
supported or not and what the plan is?

- Navid

On 5/2/14, 10:47 AM, "Luis Gomez" <ecelgp@...> wrote:

Can you please send the OSGi console printout?

On May 2, 2014, at 10:25 AM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

I just ran the latest one (0.1.2) and the GUI is not coming up at all.

Navid



On 5/1/14, 10:12 PM, "Luis Gomez" <ecelgp@...> wrote:

SP edition with latest code, you will need -Xmx2g option when starting
the controller as well.

BR/Luis


On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores
were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have
not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.



https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Cont
ro
ll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one
bridge
and bridges are controlled by controller. Overall, it is something
like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1
options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2
options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end,
it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently,
once
with 2 VMs connected to one OVS, another time with recognizing a
wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping
each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to
fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Brent Salisbury
 

Hi Navid, thanks for the interest! The GUI(s) can certainly use work. Regarding VXLAN/GRE a proposal went in yesterday to include it in the SP edition thanks to feedback such as yours.

If we can be of assistance in getting your ideal environment together always make time so please don't hesitate.

(Disclosure I wok for RedHat and am emailing from a phone that doesn't have tokens for corporate email)

Hit us up with a convenient time for you or pop into the #opendaylight-ovsdb if we can assist with the overlay components.

Thanks!
Brent

On May 2, 2014, at 7:53 PM, "Ghazisaidi, Navid" <navid.ghazisaidi@...> wrote:

Finally, I was able to bring up the GUI. It is not very stable. Sometimes
come up sometimes not….

Btw do you know who then knows if GRE and VXLAN are supposed to be
supported or not and what the plan is?

- Navid



On 5/2/14, 10:47 AM, "Luis Gomez" <ecelgp@...> wrote:

Can you please send the OSGi console printout?

On May 2, 2014, at 10:25 AM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

I just ran the latest one (0.1.2) and the GUI is not coming up at all.

Navid



On 5/1/14, 10:12 PM, "Luis Gomez" <ecelgp@...> wrote:

SP edition with latest code, you will need -Xmx2g option when starting
the controller as well.

BR/Luis


On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores
were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have
not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.



https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Cont
ro
ll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one
bridge
and bridges are controlled by controller. Overall, it is something
like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1
options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2
options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end,
it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently,
once
with 2 VMs connected to one OVS, another time with recognizing a
wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping
each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to
fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Luis Gomez
 

Hi Navid, I just made a few tests myself and new OF plugin (-of13 option) has some issues after datastore changes this week, in all editions but specially on SP and Virtualization so I do not recommend to use this option until these problems get fixed.

On the other hand OVSDB NB API should be all operational in SP edition as the only extra bundle in Virtualization edition is the OVSDB Neutron implementation.

BR/Luis

On May 3, 2014, at 11:20 AM, Brent Salisbury <brent.salisbury@...> wrote:

Hi Navid, thanks for the interest! The GUI(s) can certainly use work. Regarding VXLAN/GRE a proposal went in yesterday to include it in the SP edition thanks to feedback such as yours.

If we can be of assistance in getting your ideal environment together always make time so please don't hesitate.

(Disclosure I wok for RedHat and am emailing from a phone that doesn't have tokens for corporate email)

Hit us up with a convenient time for you or pop into the #opendaylight-ovsdb if we can assist with the overlay components.

Thanks!
Brent

Sent from my iPhone

On May 2, 2014, at 7:53 PM, "Ghazisaidi, Navid" <navid.ghazisaidi@...> wrote:

Finally, I was able to bring up the GUI. It is not very stable. Sometimes
come up sometimes not….

Btw do you know who then knows if GRE and VXLAN are supposed to be
supported or not and what the plan is?

- Navid



On 5/2/14, 10:47 AM, "Luis Gomez" <ecelgp@...> wrote:

Can you please send the OSGi console printout?

On May 2, 2014, at 10:25 AM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

I just ran the latest one (0.1.2) and the GUI is not coming up at all.

Navid



On 5/1/14, 10:12 PM, "Luis Gomez" <ecelgp@...> wrote:

SP edition with latest code, you will need -Xmx2g option when starting
the controller as well.

BR/Luis


On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3 cores
were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I have
not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.



https://wiki.opendaylight.org/view/CrossProject:Integration_Group:Cont
ro
ll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server are
connected to each other via veth. Each VM is interfacing with one
bridge
and bridges are controlled by controller. Overall, it is something
like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1
options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2
options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the end,
it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently. I
restarted controller and every time it learns network differently,
once
with 2 VMs connected to one OVS, another time with recognizing a
wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot ping
each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any plan to
fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs


Ghazisaidi, Navid <navid.ghazisaidi@...>
 

Another question: I noticed there is no any database running in controller
level to keep and restore information. I created a LISP between 2 LISPmobs
(worked well!). Then I killed the controller, packets were transmitted and
received between the ends as long as the path was in the cache. Then I
restarted the controller. The controller came up completely, the packets
didn’t get delivered. So I was forced to destroy the LISP in each end and
reregister the keys in the controller to reestablish the LISP between
ends. Shouldn’t controller at least keep the keys somewhere that in case
the controller crashed or rebooted it can continue its work? I think it
should be considered as a high-priority requirement for LISP. Is there any
plan to add DB to the controller at all?

Thanks,
Navid

On 5/3/14, 11:06 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid, I just made a few tests myself and new OF plugin (-of13 option)
has some issues after datastore changes this week, in all editions but
specially on SP and Virtualization so I do not recommend to use this
option until these problems get fixed.

On the other hand OVSDB NB API should be all operational in SP edition as
the only extra bundle in Virtualization edition is the OVSDB Neutron
implementation.

BR/Luis


On May 3, 2014, at 11:20 AM, Brent Salisbury <brent.salisbury@...>
wrote:

Hi Navid, thanks for the interest! The GUI(s) can certainly use work.
Regarding VXLAN/GRE a proposal went in yesterday to include it in the SP
edition thanks to feedback such as yours.

If we can be of assistance in getting your ideal environment together
always make time so please don't hesitate.

(Disclosure I wok for RedHat and am emailing from a phone that doesn't
have tokens for corporate email)

Hit us up with a convenient time for you or pop into the
#opendaylight-ovsdb if we can assist with the overlay components.

Thanks!
Brent

Sent from my iPhone

On May 2, 2014, at 7:53 PM, "Ghazisaidi, Navid"
<navid.ghazisaidi@...> wrote:

Finally, I was able to bring up the GUI. It is not very stable.
Sometimes
come up sometimes not….

Btw do you know who then knows if GRE and VXLAN are supposed to be
supported or not and what the plan is?

- Navid



On 5/2/14, 10:47 AM, "Luis Gomez" <ecelgp@...> wrote:

Can you please send the OSGi console printout?

On May 2, 2014, at 10:25 AM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

I just ran the latest one (0.1.2) and the GUI is not coming up at
all.

Navid



On 5/1/14, 10:12 PM, "Luis Gomez" <ecelgp@...> wrote:

SP edition with latest code, you will need -Xmx2g option when
starting
the controller as well.

BR/Luis


On May 1, 2014, at 10:03 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Thanks for your advices Luis!

- Yes I am using the latest release (forgot to mention in my
previous
email that I use Service Provider edition).

- I used -of13 option and ran it on 4 cores server and I saw 3
cores
were
in used by controller.

Regards,
Navid


On 5/1/14, 9:56 PM, "Luis Gomez" <ecelgp@...> wrote:

Hi Navid,

For the OVS features like GRE or VXLAN I cannot help much as I
have
not
tested them. Regarding controller performance a few advices:

- Use latest controller distribution, i.e. the hydrogen release
distributions had some performance issues.




https://wiki.opendaylight.org/view/CrossProject:Integration_Group:C
ont
ro
ll
er_Artifacts

- If you are using new OF plugin (-of13 option) we recommend to
use 4
Cores

Let me know if this helps

BR/Luis



On Apr 30, 2014, at 6:15 PM, Ghazisaidi, Navid
<navid.ghazisaidi@...> wrote:

Hi,

I configured a very simple scenario, where two VMs on a server
are
connected to each other via veth. Each VM is interfacing with one
bridge
and bridges are controlled by controller. Overall, it is
something
like
this: VM1‹br1 (OVS1)‹br2 (OVS2)‹VM2.

I Created GRE tunnel between two bridges using:
ovs-vsctl add-port br1 gre1 -- set interface gre1 type=gre
options:remote_ip=1.1.1.2 options:local_ip=1.1.1.1
options:egress_port=2
ovs-vsctl add-port br2 gre2 -- set interface gre2 type=gre
options:remote_ip=1.1.1.1 options:local_ip=1.1.1.2
options:egress_port=2

Here is my observations:
- CPU usage for controller instance increases rapidly. At the
end,
it
mostly requires 2 cores (sometimes 3 cores).
- Network topology is not discovered correctly and consistently.
I
restarted controller and every time it learns network
differently,
once
with 2 VMs connected to one OVS, another time with recognizing a
wrong
number of ports, and so on.
- No traffic is passing through the GRE tunnel and VMs cannot
ping
each
other, even with installing a correct set of flow entries.

Note 1. I also tested patch instead of veth and had the same
observations.
Note 2. I tested the same topology with VXLAN and had the same
observations.

I think these are the bugs and my question is GRE and VXLAN are
supposed to be supported at all or not? If yes, is there any
plan to
fix
these bugs? if not, is there any plan to support them?

Thanks,
Navid
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs
_______________________________________________
demos-and-pocs mailing list
demos-and-pocs@...
https://lists.opendaylight.org/mailman/listinfo/demos-and-pocs