Hi Chore, By default interface stats are collected for rx, tx, drops etc. However, to collect stats related to whether the RX/TX packet is unicast, multicast or broadcast requires extra checks in the data-plane that do not come for free. So, to able these extra stats, and hence incur the extra cost, is an optional feature.
Regards, neale De : <vpp-dev@lists.fd.io> au nom de emma sdi <s3m2e1.6s...@gmail.com> Date : lundi 24 décembre 2018 à 05:44 À : Ole Troan <otr...@employees.org> Cc : Marco Varlese <mvarl...@suse.de>, vpp-dev <vpp-dev@lists.fd.io> Objet : Re: [vpp-dev] Difference between python api and vppctl result - stats There is 'stats-collect-rx' and 'stats-collect-tx' nodes that increment unicast, multicat and broadcast counters. these 2 nodes are not registered by default. So you have to use 'collect_detailed_interface_stats' API to register these 2 nodes in order to start these combined counters to count. Best Regards Chore On Sun, Dec 23, 2018 at 9:24 AM Ole Troan <otr...@employees.org<mailto:otr...@employees.org>> wrote: Hi there, On 23 Dec 2018, at 13:53, emma sdi <s3m2e1.6s...@gmail.com<mailto:s3m2e1.6s...@gmail.com>> wrote: Hi Dear Ole I used vpp_papi.vpp_stats but still it is necessary to call "collect_detailed_interface_stats API" to collect rx-unicast, tx-unicast and etc. why the detailed_interface_stats collector node is not available by default? Ok, good. I am not quite sure what you mean by detailed_interface_stats collector node. Could you elaborate/point to code? Cheers Ole On Sun, Dec 23, 2018 at 8:11 AM Ole Troan <otr...@employees.org<mailto:otr...@employees.org>> wrote: Hello, Please use the new stats api. In Python that’s available in vpp_papi.vpp_stats. The old stats api over the vpp binary api (stats.api) is deprecated. Let me know if you need help. Cheers Ole On 23 Dec 2018, at 12:50, khers <s3m2e1.6s...@gmail.com<mailto:s3m2e1.6s...@gmail.com>> wrote: Hi Dear VPP I was testing interface stats in VPP 18.10. It seems that it is needed to use collect_detailed_interface_stats API for every interface to register collector node. As a result, all combined counters such as unicast and broadcast start to work. My question is, why this collector is not enabled by default? how does it affect throughput? On Tue, Sep 18, 2018 at 1:44 AM Marco Varlese <mvarl...@suse.de<mailto:mvarl...@suse.de>> wrote: On Mon, 2018-09-17 at 20:33 +0200, Ole Troan wrote: > Marco, > > > > > Which version of vpp python api does include this stats api? > > > > I tried it at vpp stable/1807, but there was no VPPStats module in it! > > > > > > Going in 18.10. I recommend latest master for now. > > > > Great! Any chances to see it added to the wiki page for 18.10? > > https://wiki.fd.io/view/Projects/vpp/Release_Plans/Release_Plan_18.10 > > Now there is a probability near 1 that you will see it on the wiki page! ;-) Thanks :) > > Best regards, > Ole > > > > > > > > > On Mon, Sep 17, 2018 at 11:45 AM Ole Troan > > > > <otr...@employees.org<mailto:otr...@employees.org>> wrote: > > > > > Brayan, > > > > > > > > > > > I would like to share my problem with you in order to find solution. > > > > > > As you can see in vppctl output file, "vppctl show hardware" command > > > > > > result contains "rx broadcast packets" and "tx broadcast packets". > > > > > > However, when I used vpp python api to extract these stats, the > > > > > > value of VNET_INTERFACE_COUNTER_RX_BROADCAST and > > > > > > VNET_INTERFACE_COUNTER_TX_BROADCAST are stuck on zero incorrectly. > > > > > > attached you can see my python script and its output and vppctl > > > > > > command output. This behavior exists in multicast stats. > > > > > > > > > > > > Best Regards, > > > > > > > > > > > > > > > > > > vppctl output: https://pastebin.com/zXrArE4b > > > > > > Python Script: https://pastebin.com/gDJbH1if > > > > > > Python Script output: https://pastebin.com/q4dJ8VMs > > > > > > > > > > Firstly these are not exactly the same counters. The first ones come > > > > > directly from DPDK I believe, and the VNET_INTERFACE_COUNTER ones are > > > > > from the stats_collect_rx_node, sitting as an input feature behind > > > > > device-input. > > > > > I don’t know if they are different because the broadcasts are filtered > > > > > out before reaching VPP or if there is a better explanation, Damjan? > > > > > > > > > > With regards to Python and stats collection, or any stats collection > > > > > in general. I would encourage you to try out the new stats segment. > > > > > > > > > > In Python you basically do: > > > > > > > > > > from vpp_papi.vpp_stats import VPPStats > > > > > stat = VPPStats(<name-of-stats-socket) > > > > > dir = stat.ls<http://stat.ls>(['^/if’]) # For all interface counters > > > > > use “show > > > > > statistics segment” or vpp_get_stats ls to see all available counters > > > > > counters = stat.dump(dir) > > > > > > > > > > Then counters[‘/if/tx'][thread index][sw_if_index] will give you the > > > > > TX counter for a particular worker thread / software interface. > > > > > > > > > > Cheers, > > > > > Ole > > > > > > > > > > > > > > > > > > > > > > > -=-=-=-=-=-=-=-=-=-=-=- > > > > Links: You receive all messages sent to this group. > > > > > > > > View/Reply Online (#10515): https://lists.fd.io/g/vpp-dev/message/10515 > > > > Mute This Topic: https://lists.fd.io/mt/25704565/675193 > > > > Group Owner: > > > > vpp-dev+ow...@lists.fd.io<mailto:vpp-dev%2bow...@lists.fd.io> > > > > Unsubscribe: https://lists.fd.io/g/vpp-dev/unsubb > > > > [otr...@employees.org<mailto:otr...@employees.org> > > > > ] > > > > -=-=-=-=-=-=-=-=-=-=-=- > > > > > > -=-=-=-=-=-=-=-=-=-=-=- > > > Links: You receive all messages sent to this group. > > > > > > View/Reply Online (#10516): https://lists.fd.io/g/vpp-dev/message/10516 > > > Mute This Topic: https://lists.fd.io/mt/25704565/675056 > > > Group Owner: vpp-dev+ow...@lists.fd.io<mailto:vpp-dev%2bow...@lists.fd.io> > > > Unsubscribe: https://lists.fd.io/g/vpp-dev/unsubb > > > [mvarl...@suse.de<mailto:mvarl...@suse.de>] > > > -=-=-=-=-=-=-=-=-=-=-=- > > > > > > > -- > > Marco V > > > > SUSE LINUX GmbH | GF: Felix Imendörffer, Jane Smithard, Graham Norton > > HRB 21284 (AG Nürnberg) Maxfeldstr. 5, D-90409, Nürnberg > > -=-=-=-=-=-=-=-=-=-=-=- > Links: You receive all messages sent to this group. > > View/Reply Online (#10527): https://lists.fd.io/g/vpp-dev/message/10527 > Mute This Topic: https://lists.fd.io/mt/25704565/675056 > Group Owner: vpp-dev+ow...@lists.fd.io<mailto:vpp-dev%2bow...@lists.fd.io> > Unsubscribe: https://lists.fd.io/g/vpp-dev/unsubb > [mvarl...@suse.de<mailto:mvarl...@suse.de>] > -=-=-=-=-=-=-=-=-=-=-=- -- Marco V SUSE LINUX GmbH | GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg) Maxfeldstr. 5, D-90409, Nürnberg -=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#10530): https://lists.fd.io/g/vpp-dev/message/10530 Mute This Topic: https://lists.fd.io/mt/25704565/675776 Group Owner: vpp-dev+ow...@lists.fd.io<mailto:vpp-dev%2bow...@lists.fd.io> Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [s3m2e1.6s...@gmail.com<mailto:s3m2e1.6s...@gmail.com>] -=-=-=-=-=-=-=-=-=-=-=- -=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#11767): https://lists.fd.io/g/vpp-dev/message/11767 Mute This Topic: https://lists.fd.io/mt/25704565/675193 Group Owner: vpp-dev+ow...@lists.fd.io<mailto:vpp-dev+ow...@lists.fd.io> Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [otr...@employees.org<mailto:otr...@employees.org>] -=-=-=-=-=-=-=-=-=-=-=-
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#11771): https://lists.fd.io/g/vpp-dev/message/11771 Mute This Topic: https://lists.fd.io/mt/25704565/21656 Group Owner: vpp-dev+ow...@lists.fd.io Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
