Re: [vpp-dev] Multiple VRFs in 1609

[Michael Borokhovich]

Hi Florin,

Thank you for the response. The reason I stick to 1609 is the router plugin
(vppsb) that works with 1609 but does not with any other VPP version I
tried. So, I'm trying to get multiple VRFs works with 1609 since I also
need the router plugin for the project...

Thanks,
Michael.


On Wed, Oct 4, 2017 at 10:33 PM, Florin Coras 
wrote:

> Michael,
>
> I would recommend you switched to a newer release (17.07 or the soon to be
> release 17.10) since the fib code has been completely reworked in 17.01.
>
> Florin
>
> On Oct 4, 2017, at 7:13 PM, Michael Borokhovich 
> wrote:
>
> Hi,
>
> I'm trying to configure the following setup.
>
> GigabitEthernet0/4/0 - Table 1
> GigabitEthernet0/5/0 - Table 2
> GigabitEthernet0/6/0 - Table 0 (default)
>
> If a packet with DST_IP="10.5.1.0/24" received at GigabitEthernet0/4/0 or
> GigabitEthernet0/5/0 it needs to be sent via GigabitEthernet0/6/0.
>
> The following works for VPP 1710:
>
> vppctl ip route add 10.5.1.0/24 table 1 via 10.5.4.11 GigabitEthernet0/6/0
>
> vppctl ip route add 10.5.1.0/24 table 2 via 10.5.4.11 GigabitEthernet0/6/0
>
> But for 1609 it does not work. The routes are added to the default table 0
> instead of tables 1 and 2.
>
> If I do just this (just interface name, without the next ho IP):
>
> vppctl ip route add 10.5.1.0/24 table 1 via GigabitEthernet0/6/0
>
> Then the route is installed correctly to table 1 but the outgoing packets
> are messed up - sent without ethernet header.
>
>
> Any ideas how to make multiple VRFs woks in 1609?
>
>
> Thanks,
>
> Michael.
>
>
> ___
> vpp-dev mailing list
> vpp-dev@lists.fd.io
> https://lists.fd.io/mailman/listinfo/vpp-dev
>
>
>
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] Multiple VRFs in 1609

[Florin Coras]

Michael, 

I would recommend you switched to a newer release (17.07 or the soon to be 
release 17.10) since the fib code has been completely reworked in 17.01. 

Florin

> On Oct 4, 2017, at 7:13 PM, Michael Borokhovich  wrote:
> 
> Hi,
> 
> I'm trying to configure the following setup.
> 
> GigabitEthernet0/4/0 - Table 1
> GigabitEthernet0/5/0 - Table 2
> GigabitEthernet0/6/0 - Table 0 (default)
> 
> If a packet with DST_IP="10.5.1.0/24 " received at 
> GigabitEthernet0/4/0 or GigabitEthernet0/5/0 it needs to be sent via 
> GigabitEthernet0/6/0.
> 
> The following works for VPP 1710:
> vppctl ip route add 10.5.1.0/24  table 1 via 10.5.4.11 
> GigabitEthernet0/6/0
> 
> 
> vppctl ip route add 10.5.1.0/24  table 2 via 10.5.4.11 
> GigabitEthernet0/6/0
> 
> But for 1609 it does not work. The routes are added to the default table 0 
> instead of tables 1 and 2.
> 
> If I do just this (just interface name, without the next ho IP): 
> 
> vppctl ip route add 10.5.1.0/24  table 1 via 
> GigabitEthernet0/6/0
> 
> Then the route is installed correctly to table 1 but the outgoing packets are 
> messed up - sent without ethernet header.
> 
> 
> 
> Any ideas how to make multiple VRFs woks in 1609?
> 
> 
> 
> Thanks,
> 
> Michael.
> 
> 
> 
> ___
> vpp-dev mailing list
> vpp-dev@lists.fd.io
> https://lists.fd.io/mailman/listinfo/vpp-dev

___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

[vpp-dev] Multiple VRFs in 1609

[Michael Borokhovich]

Hi,

I'm trying to configure the following setup.

GigabitEthernet0/4/0 - Table 1
GigabitEthernet0/5/0 - Table 2
GigabitEthernet0/6/0 - Table 0 (default)

If a packet with DST_IP="10.5.1.0/24" received at GigabitEthernet0/4/0 or
GigabitEthernet0/5/0 it needs to be sent via GigabitEthernet0/6/0.

The following works for VPP 1710:

vppctl ip route add 10.5.1.0/24 table 1 via 10.5.4.11 GigabitEthernet0/6/0

vppctl ip route add 10.5.1.0/24 table 2 via 10.5.4.11 GigabitEthernet0/6/0

But for 1609 it does not work. The routes are added to the default table 0
instead of tables 1 and 2.

If I do just this (just interface name, without the next ho IP):

vppctl ip route add 10.5.1.0/24 table 1 via GigabitEthernet0/6/0

Then the route is installed correctly to table 1 but the outgoing packets
are messed up - sent without ethernet header.


Any ideas how to make multiple VRFs woks in 1609?


Thanks,

Michael.
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] Issue with tagged packets on DPDK-interface across VMs

[Balaji Kn]

Hi All,

I tried with both uio_pci_generic driver and igb_uio driver. Please can you
share your opinion on this?

Regards,
Balaji

On Tue, Oct 3, 2017 at 5:59 PM, Balaji Kn  wrote:

> Hi All,
>
> I am working on VPP 17.07 and using ubuntu 14.04. I have two VMs say VM1
> and VM2. I am running VPP on VM1 and interface between VM1 and VM2 is DPDK
> type.
>
> *Configuration*
> vpp# set int state GigabitEthernet0/a/0 up
>
> vpp#
>
> vpp# create sub-interface GigabitEthernet0/a/0 500 dot1q 500 exact-match
> GigabitEthernet0/a/0.500
>
> vpp#
> vpp# set int ip address GigabitEthernet0/a/0.500 5.5.5.1/24
>
> vpp# set interface state GigabitEthernet0/a/0.500 up
>
> vpp# show int addr
> GigabitEthernet0/10/0 (up):
> GigabitEthernet0/9/0 (up):
> GigabitEthernet0/a/0 (up):
> GigabitEthernet0/a/0.500 (up):
>   5.5.5.1/24
> local0 (dn):
>
>
> Tagged packets  received on VPP are dropped as l3 mac mismatch. Enabling
> trace is showing all packets received with incorrect hardware address.
>
> *Trace*
> 00:13:14:020654: dpdk-input
>   GigabitEthernet0/a/0 rx queue 0
>   buffer 0xe487: current data 0, length 64, free-list 0, clone-count 0,
> totlen-nifb 0, trace 0x4
>   PKT MBUF: port 1, nb_segs 1, pkt_len 64
> buf_len 2176, data_len 64, ol_flags 0x0, data_off 128, phys_addr
> 0x7514ac40
> packet_type 0x0
>   0x: 00:00:00:00:00:00 -> 00:00:00:00:00:00
> 00:13:14:020708: ethernet-input
>   0x: 00:00:00:00:00:00 -> 00:00:00:00:00:00
> 00:13:14:020730: error-drop
>   ethernet-input: l3 mac mismatch
>
> If i ping from VPP,  VM2 is responding with ARP reply. Is there any known
> issue with tagged packets for DPDK interfaces across VMs? Any pointers
> would be appreciated.
>
> However there are no issues with base interface on same setup. I am not
> sure whether tagged packets are corrupted or not on dpdk-input.
>
> Regards,
> Balaji
>
>
>
>
>
>
>
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] Running multiple instances of VPP as it was allowed with VPP-lite

[Florin Coras]

Hi Justin, 

Try api-segment {prefix }

Florin

> On Oct 4, 2017, at 9:23 AM, Justin Iurman  wrote:
> 
> Hi all,
> 
> Is it still possible to run multiple instances of VPP, just like it was done 
> with VPP-lite (see here: 
> https://wiki.fd.io/view/VPP/Progressive_VPP_Tutorial) before merging it ?
> 
> Actually, my problem is the following. I need to run several instances of VPP 
> (eg. vpp1, vpp2, vpp3, etc) to simulate a topology. But, for some reason, I'm 
> not able to make it work as expected. I've already disabled DPDK plugin to 
> run in "lite" mode. Could it be a configuration issue ? VPP startup.conf:
> 
> unix {
>  nodaemon
>  log /tmp/vpp.log
>  full-coredump
>  cli-listen /run/vpp/cli.sock
>  gid vpp
> }
> 
> api-trace {
>  on
> }
> 
> api-segment {
>  gid vpp
> }
> 
> plugins {
>  plugin dpdk_plugin.so { disable }
> }
> 
> I tried to compare both VPP and VPP-lite. A first observation showed that VPP 
> is installed as a service, which is not the case for VPP-lite. I've also 
> compared both config files. Here is VPP-lite startup.conf:
> 
> unix {
>  nodaemon
>  log /tmp/vpp.log
>  full-coredump
> }
> 
> api-trace {
>  on
> }
> 
> api-segment {
>  gid vpp
> }
> 
> Currently, here is how I try to mimic VPP-lite behavior with VPP:
> 
> sudo vpp api-segment { prefix vpp1 } unix { cli-listen /run/vpp/cli.vpp1.sock 
> }
> sudo vppctl -s /run/vpp/cli.vpp1.sock
> 
> It works for one instance, but not for more, even if it does not throw any 
> error. Do you have any idea on how to proceed or what is wrong here ?
> 
> Thanks !
> 
> Justin
> ___
> vpp-dev mailing list
> vpp-dev@lists.fd.io
> https://lists.fd.io/mailman/listinfo/vpp-dev

___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

[vpp-dev] Running multiple instances of VPP as it was allowed with VPP-lite

[Justin Iurman]

Hi all,

Is it still possible to run multiple instances of VPP, just like it was done 
with VPP-lite (see here: https://wiki.fd.io/view/VPP/Progressive_VPP_Tutorial) 
before merging it ?

Actually, my problem is the following. I need to run several instances of VPP 
(eg. vpp1, vpp2, vpp3, etc) to simulate a topology. But, for some reason, I'm 
not able to make it work as expected. I've already disabled DPDK plugin to run 
in "lite" mode. Could it be a configuration issue ? VPP startup.conf:

unix {
  nodaemon
  log /tmp/vpp.log
  full-coredump
  cli-listen /run/vpp/cli.sock
  gid vpp
}

api-trace {
  on
}

api-segment {
  gid vpp
}

plugins {
  plugin dpdk_plugin.so { disable }
}

I tried to compare both VPP and VPP-lite. A first observation showed that VPP 
is installed as a service, which is not the case for VPP-lite. I've also 
compared both config files. Here is VPP-lite startup.conf:

unix {
  nodaemon
  log /tmp/vpp.log
  full-coredump
}

api-trace {
  on
}

api-segment {
  gid vpp
}

Currently, here is how I try to mimic VPP-lite behavior with VPP:

sudo vpp api-segment { prefix vpp1 } unix { cli-listen /run/vpp/cli.vpp1.sock }
sudo vppctl -s /run/vpp/cli.vpp1.sock

It works for one instance, but not for more, even if it does not throw any 
error. Do you have any idea on how to proceed or what is wrong here ?

Thanks !

Justin
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] VPP Multi-thread

[Alessio Silvestro]

Yeah you completely got the point.

Thanks for the hint ;)

Alessio

On Oct 4, 2017 12:11, "Damjan Marion"  wrote:

>
>
> On 2 Oct 2017, at 18:14, Alessio Silvestro 
> wrote:
>
> Dear all,
>
> I am running VPP on a CPU with 2 sockets and 4 virtual cores. The startup
> configuration is the following:
>
> unix {
>>   interactive
>>   nodaemon
>> }
>>
>> cpu {
>> main-core 0
>> corelist-workers 2-3
>> *workers 2*
>> }
>>
>>  dpdk {
>> dev :07:00.0
>
>{
>> num-rx-queues 2
>> }
>> dev :07:00.1
>
> socket-mem 1024,1024
>> }
>
>
> The thread placement is the following:
>
>> Thread 1 (vpp_wk_0 at lcore 2):
>>   *TenGigabitEthernet7/0/0 queue 0*
>>   TenGigabitEthernet7/0/1 queue 0
>> Thread 2 (vpp_wk_1 at lcore 3):
>>  * TenGigabitEthernet7/0/0 queue 1*
>
>
> So, I can see that RSS is working because the first queue of the first
> interface is on lcore2 whereas the second queue is on lcore3.
>
> I am running a simple L2-xconnect with the following command:
>
>> set int l2 xconnect TenGigabitEthernet7/0/0 TenGigabitEthernet7/0/1
>
>
>
> I am expecting the second thread actually working when the first cannot
> handle all the traffic.
>
> Therefore, I am saturating the RX bandwidth of TenGigabitEthernet7/0/0 sending
> 14 Mpps (packets of 64B).
> I am receiving on TenGigabitEthernet7/0/1 only ~13Mpps which means that I
> am not able to process ~ 1Mpps.
>
> However, when I perform  "show run", the second thread does not work at
> all!
>
> Am I missing something in the configuration of VPP or what do you think
> can be the cause?
>
> Thanks for the help :)
>
>
> RSS is per-flow so likely your NIC thinks that all your traffic belongs to
> the same flow and sends it to the same queue.
> You need to make your traffic more diverse, i.e. by randomizing address or
> port values…
>
>
>
>
>
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] [vpp-v17.10] L2 forwarding errors in vhost-user configuration

[John Lo (loj)]

The "proper" CLI is "show node counters" which displays stats kept by various 
graph nodes including normal operation counts and possibly error counts. The 
CLI "sho err" is historical and was kept for backward compatibilities. The old 
CLI is somewhat misleading but is quicker to type then the proper one.   -John

From: vpp-dev-boun...@lists.fd.io [mailto:vpp-dev-boun...@lists.fd.io] On 
Behalf Of Saxena, Nitin
Sent: Wednesday, October 04, 2017 10:52 AM
To: Damjan Marion 
Cc: Athreya, Narayana Prasad ; 
vpp-dev@lists.fd.io
Subject: Re: [vpp-dev] [vpp-v17.10] L2 forwarding errors in vhost-user 
configuration


Thanks for the clarification Damjan



-Nitin


From: Damjan Marion >
Sent: Wednesday, October 4, 2017 3:37 PM
To: Saxena, Nitin
Cc: vpp-dev@lists.fd.io; Steven Luong (sluong); 
Athreya, Narayana Prasad
Subject: Re: [vpp-dev] [vpp-v17.10] L2 forwarding errors in vhost-user 
configuration






On 3 Oct 2017, at 17:32, Saxena, Nitin 
> wrote:

Hi,

While running ping between VM1 to VM2 on aarch64, configured using vhost-user 
interface, I am seeing following L2 errors in VPP v1710 for each ICMP packet 
sent


DBGvpp# show err
   CountNode  Reason
  1096l2-output   L2 output packets
  1096l2-learnL2 learn packets
 2l2-learnL2 learn misses
36l2-learnL2 learn hit updates
  1096l2-inputL2 input packets
35l2-floodL2 flood packets
 8 VirtualEthernet0/0/0-txtx packet drops (no available 
descriptors)
DBGvpp# sh int
  Name   Idx   State  Counter  Count
VirtualEthernet0/0/0  1 up   rx packets 
  552
 rx bytes   
52876
 tx packets 
  552
 tx bytes   
52876
 drops  
8
VirtualEthernet0/0/1  2 up   rx packets 
  552
 rx bytes   
52876
 tx packets 
  552
 tx bytes   
52876
local00down
DBGvpp# show version
vpp v17.10-rc1~4-gbd70c2f built by nsaxena on sbr2s-10 at Tue Oct  3 08:16:41 
UTC 2017


Are these errors are false alarm ? While ping is working fine between two VM's 
these errors are raising false alarms to me. Seems like a problem to me. If 
something to be fixed I can contribute to fix these errors.

Those counters are simply counting packets in the l2fib. You will get same 
output if  you do "show counters".
nothing to worry about. Still I agree that this is confusing and that we might 
consider separating error counters from the normal ones...


___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

[vpp-dev] Packet data and vlib_buffer contiguity

[Francois Ozog]

Hi,

Hardware that is capable of 50Gbps and above (at 64 byte line rate)
place packets next to each other in large memory zones rather than in
individual memory buffers.

To handle packets without copy would require vlib_buffer_t to allow
packet data to be NOT consecutive to it.

Are there plans or have there been discussions on this topic before ?
(I checked the archive and did not found any reference).

-FF
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

[vpp-dev] Fixes not being cherry picked to stable/1710

[Florin Coras]

Folks, 

I see a number of patches on master that look like fixes, have the word “fix” 
in subject line, but no jira ticket assigned and have not been cherry-picked to 
stable/1710. Could patch owners, once their patches have been merged, take care 
of these last two steps *if needed* and thereby avoid possible future pain to 
be created by me haphazardly porting their fixes? Also, in closing, let me 
remind you all the wise teachings of our creed:

• All bug fixes must be double-committed to the release throttle as 
well as to the master branch
• Commit first to the release throttle, then "git cherry-pick" 
into master
• Manual merges may be required, depending on the degree of 
divergence between throttle and master
• All bug fixes need to have a Jira ticket
• Please put Jira IDs into the commit messages.
• Please use the same Jira ID

Regards, 
Florin
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] [vpp-v17.10] L2 forwarding errors in vhost-user configuration

[Saxena, Nitin]

Thanks for the clarification Damjan


-Nitin


From: Damjan Marion 
Sent: Wednesday, October 4, 2017 3:37 PM
To: Saxena, Nitin
Cc: vpp-dev@lists.fd.io; Steven Luong (sluong); Athreya, Narayana Prasad
Subject: Re: [vpp-dev] [vpp-v17.10] L2 forwarding errors in vhost-user 
configuration





On 3 Oct 2017, at 17:32, Saxena, Nitin 
> wrote:


Hi,

While running ping between VM1 to VM2 on aarch64, configured using vhost-user 
interface, I am seeing following L2 errors in VPP v1710 for each ICMP packet 
sent


DBGvpp# show err
   CountNode  Reason
  1096l2-output   L2 output packets
  1096l2-learnL2 learn packets
 2l2-learnL2 learn misses
36l2-learnL2 learn hit updates
  1096l2-inputL2 input packets
35l2-floodL2 flood packets
 8 VirtualEthernet0/0/0-txtx packet drops (no available 
descriptors)
DBGvpp# sh int
  Name   Idx   State  Counter  Count
VirtualEthernet0/0/0  1 up   rx packets 
  552
 rx bytes   
52876
 tx packets 
  552
 tx bytes   
52876
 drops  
8
VirtualEthernet0/0/1  2 up   rx packets 
  552
 rx bytes   
52876
 tx packets 
  552
 tx bytes   
52876
local00down
DBGvpp# show version
vpp v17.10-rc1~4-gbd70c2f built by nsaxena on sbr2s-10 at Tue Oct  3 08:16:41 
UTC 2017


Are these errors are false alarm ? While ping is working fine between two VM's 
these errors are raising false alarms to me. Seems like a problem to me. If 
something to be fixed I can contribute to fix these errors.

Those counters are simply counting packets in the l2fib. You will get same 
output if  you do “show counters”.
nothing to worry about. Still I agree that this is confusing and that we might 
consider separating error counters from the normal ones...


___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] [releng] Proposal to redirect #opendaylight-releng to #lf-releng

[Ed Warnicke]

One  suggestion I would have would be to do a phased transition, with a
'topic' in the existing infra channels set to direct folks to lf-releng,
and a period (3 months?, 6 months?) of keeping a presence on the existing
channels, to be used  to politely request bouncing conversations that arise
in them to lf-releng... this way folks can get used to it, and in worst
case, see in the topic where to go :)

Ed

On Tue, Oct 3, 2017 at 11:12 AM, Thanh Ha 
wrote:

> On Tue, Sep 26, 2017 at 6:38 PM, Thanh Ha 
> wrote:
>
>> Hi Everyone,
>>
>> We'd like to pitch an idea to have #opendaylight-releng irc channel
>> redirect to a new #lf-releng channel. Something that's occurred to us is
>> that many of the networking at LF projects have their own separate releng
>> channels in which folks typically ask JJB related questions. Each of these
>> channels are typically moderately active.
>>
>> Something we've been thinking about is the idea of merging all these
>> releng channels into #lf-releng which we're hoping can combine the
>> communities JJB experts so that JJB and other releng related questions can
>> be more broadly asked.
>>
>> Thoughts?
>>
>> Regards,
>> Thanh
>>
>
> Hi Everyone,
>
> For those that don't know me. I am one of the release engineers working on
> the OpenDaylight project. I've added the fd.io, opnfv, and onap
> communities to this list to get feedback on the idea proposed above. Let me
> know if there's other mailing lists in the respective projects we should be
> cc'ing to include in the discussions.
>
> I'd like to hear from the fd.io, opnfv, and onap for their thoughts on
> the proposal to create a single lf-releng channel on IRC and have all the
> respective releng channels redirect to it. Since our respective releng
> projects use similar technologies like JJB, Jenkins, etc... it might be a
> benefit to pool together our expertise so that it is easier to ask
> questions to the broader community when we need help with things like JJB,
> Gerrit, etc...
>
> You can follow the discussion thread so far from the ODL community
> discussed here:
> https://lists.opendaylight.org/pipermail/dev/2017-September/004066.html
>
>
> Regards,
> Thanh
>
>
> ___
> vpp-dev mailing list
> vpp-dev@lists.fd.io
> https://lists.fd.io/mailman/listinfo/vpp-dev
>
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] [releng] Proposal to redirect #opendaylight-releng to #lf-releng

[Thomas F Herbert]

+ci-...@fd.io

This probably should also go to ci-man which is roubhly the equivalent 
of releng in opnfv.


--TFH

On 10/03/2017 02:12 PM, Thanh Ha wrote:
On Tue, Sep 26, 2017 at 6:38 PM, Thanh Ha 
> 
wrote:


Hi Everyone,

We'd like to pitch an idea to have #opendaylight-releng irc
channel redirect to a new #lf-releng channel. Something that's
occurred to us is that many of the networking at LF projects have
their own separate releng channels in which folks typically ask
JJB related questions. Each of these channels are typically
moderately active.

Something we've been thinking about is the idea of merging all
these releng channels into #lf-releng which we're hoping can
combine the communities JJB experts so that JJB and other releng
related questions can be more broadly asked.

Thoughts?

Regards,
Thanh


Hi Everyone,

For those that don't know me. I am one of the release engineers 
working on the OpenDaylight project. I've added the fd.io 
, opnfv, and onap communities to this list to get 
feedback on the idea proposed above. Let me know if there's other 
mailing lists in the respective projects we should be cc'ing to 
include in the discussions.


I'd like to hear from the fd.io , opnfv, and onap for 
their thoughts on the proposal to create a single lf-releng channel on 
IRC and have all the respective releng channels redirect to it. Since 
our respective releng projects use similar technologies like JJB, 
Jenkins, etc... it might be a benefit to pool together our expertise 
so that it is easier to ask questions to the broader community when we 
need help with things like JJB, Gerrit, etc...


You can follow the discussion thread so far from the ODL community 
discussed here:

https://lists.opendaylight.org/pipermail/dev/2017-September/004066.html


Regards,
Thanh



___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev


--
*Thomas F Herbert*
NFV and Fast Data Planes
Office of Technology
*Red Hat*
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] VPP Multi-thread

[Damjan Marion]

> On 2 Oct 2017, at 18:14, Alessio Silvestro  wrote:
> 
> Dear all,
> 
> I am running VPP on a CPU with 2 sockets and 4 virtual cores. The startup 
> configuration is the following:
> 
> unix {
>   interactive
>   nodaemon
> }
> 
> cpu {
> main-core 0
>   corelist-workers 2-3
>   workers 2
> }
> 
>  dpdk {
>   dev :07:00.0 
>{
>   num-rx-queues 2
>   }
>   dev :07:00.1
>   socket-mem 1024,1024
> }
> 
> The thread placement is the following:
> Thread 1 (vpp_wk_0 at lcore 2):
>   TenGigabitEthernet7/0/0 queue 0
>   TenGigabitEthernet7/0/1 queue 0
> Thread 2 (vpp_wk_1 at lcore 3):
>   TenGigabitEthernet7/0/0 queue 1
> 
> So, I can see that RSS is working because the first queue of the first 
> interface is on lcore2 whereas the second queue is on lcore3.
> 
> I am running a simple L2-xconnect with the following command:
> set int l2 xconnect TenGigabitEthernet7/0/0 TenGigabitEthernet7/0/1
> 
>  
> I am expecting the second thread actually working when the first cannot 
> handle all the traffic. 
> 
> Therefore, I am saturating the RX bandwidth of TenGigabitEthernet7/0/0 
> sending 14 Mpps (packets of 64B).
> I am receiving on TenGigabitEthernet7/0/1 only ~13Mpps which means that I am 
> not able to process ~ 1Mpps.
> 
> However, when I perform  "show run", the second thread does not work at all!
> 
> Am I missing something in the configuration of VPP or what do you think can 
> be the cause?
> 
> Thanks for the help :)

RSS is per-flow so likely your NIC thinks that all your traffic belongs to the 
same flow and sends it to the same queue.
You need to make your traffic more diverse, i.e. by randomizing address or port 
values…




___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] [vpp-v17.10] L2 forwarding errors in vhost-user configuration

[Damjan Marion]

> On 3 Oct 2017, at 17:32, Saxena, Nitin  wrote:
> 
> Hi,
> 
> While running ping between VM1 to VM2 on aarch64, configured using vhost-user 
> interface, I am seeing following L2 errors in VPP v1710 for each ICMP packet 
> sent
> 
> 
> DBGvpp# show err
>CountNode  Reason
>   1096l2-output   L2 output packets
>   1096l2-learnL2 learn packets
>  2l2-learnL2 learn misses
> 36l2-learnL2 learn hit updates
>   1096l2-inputL2 input packets
> 35l2-floodL2 flood packets
>  8 VirtualEthernet0/0/0-txtx packet drops (no 
> available descriptors)
> DBGvpp# sh int
>   Name   Idx   State  Counter  
> Count 
> VirtualEthernet0/0/0  1 up   rx packets   
> 552
>  rx bytes 
>   52876
>  tx packets   
> 552
>  tx bytes 
>   52876
>  drops
>   8
> VirtualEthernet0/0/1  2 up   rx packets   
> 552
>  rx bytes 
>   52876
>  tx packets   
> 552
>  tx bytes 
>   52876
> local00down  
> DBGvpp# show version
> vpp v17.10-rc1~4-gbd70c2f built by nsaxena on sbr2s-10 at Tue Oct  3 08:16:41 
> UTC 2017
> 
> 
> Are these errors are false alarm ? While ping is working fine between two 
> VM's these errors are raising false alarms to me. Seems like a problem to me. 
> If something to be fixed I can contribute to fix these errors.

Those counters are simply counting packets in the l2fib. You will get same 
output if  you do “show counters”.
nothing to worry about. Still I agree that this is confusing and that we might 
consider separating error counters from the normal ones...


___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] vpp_configure_args_vpp = --disable-japi compilation issue

[Avinash Dhar Dubey]

Thanks Damjan.

On Tue, Oct 3, 2017 at 3:23 PM, Damjan Marion (damarion)  wrote:

>
>
>
> On 3 Oct 2017, at 11:47, Avinash Dhar Dubey 
> wrote:
>
> Hello,
>
> I am trying to compile vpp with flag vpp_configure_args_vpp =
> --disable-japi by modifying the file datapath/vpp/build-data/platforms/
> vpp.mk. Its resulting in broken deb packages.
>
> Any help on how to disable japi as i want to compile vpp with minimal
> dependencies.
>
>
> deb packaging doesn’t support custom configurations. If you go that way
> you will need to take care for packaging by yourself…
>
>
>
___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev

Re: [vpp-dev] Load balancer plugin for VPP

[Pierre Pfister (ppfister)]

Do you start VPP with the 'make run' command ?

It looks like the way you start VPP, it doesn't specify the righ plugins 
directory.
I'd suggest your try with adding "plugin_path " to 
VPP's command-line arguments.

- Pierre


Le 3 oct. 2017 à 20:40, John Wei 
> a écrit :

Pierre,
Thanks for the response.
I follow instructions to install VPP:
https://wiki.fd.io/view/VPP/Installing_VPP_binaries_from_packages
That gives me vpp utility. There is no lb command.
I did a build of vpp by follow the instructions below to build vpp on linux
https://wiki.fd.io/view/VPP/Pulling,_Building,_Running,_Hacking_and_Pushing_VPP_Code#Linux
I found "lb" in the plugins directory.

But, how to bring up vpp with lb plugin? I hoping to see some instructions that 
I can follow to try this VPP lb.

John


On Sun, Oct 1, 2017 at 11:55 PM, Pierre Pfister (ppfister) 
> wrote:
Hello John,

The LB plugin comes with VPP by default (no need for additional download, 
sources are in src/plugins/lb).

As far as instructions are concerned, I am afraid the link you pointed to is 
all there is for now.
But please ask if you feel like something in particular is missing.

- Pierre



Le 30 sept. 2017 à 00:07, John Wei 
> a écrit :

I saw a VPP Load Balancer mentioned in this document:

https://docs.fd.io/vpp/17.07/lb_plugin_doc.html

But, I did not see installation information.
Can someone point me to where I can get download and setup instructions?

Thanks
John

___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev



___
vpp-dev mailing list
vpp-dev@lists.fd.io
https://lists.fd.io/mailman/listinfo/vpp-dev