On 3/3/23 11:20, Olaf Seibert via dev wrote: > Tangentially related to this: I was trying to detect when the problem with > the "too many resubmits" occurs, for alerting purposes. > Some light examination of the source of ovs suggested that the coverage > counter "drop_action_too_many_resubmit" corresponds to the error > "XLATE_TOO_MANY_RESUBMITS". > > So I saw this very recent log message in > /var/log/openvswitch/ovs-vswitchd.log (I censored some addresses): > > 2023-03-03T10:06:51.453Z|00518|ofproto_dpif_xlate(handler34)|WARN|over 4096 > resubmit actions on bridge br-int while processing > arp,in_port=1,vlan_tci=0x0000,dl_src=zz:zz:zz:zz:zz:zz,dl_dst=ff:ff:ff:ff:ff:ff,arp_spa=xxx.xxx.xxx.xxx,arp_tpa=yyy.yyy.yyy.yyy,arp_op=1,arp_sha=vv:vv:vv:vv:vv:vv,arp_tha=00:00:00:00:00:00 > > but the coverage counter seems to ignore this: > > # ovs-appctl coverage/read-counter drop_action_too_many_resubmit > 0 > > Is this a bug or am I looking at the wrong thing(s)?
Coverage counters are collected only for actions executed in userspace. In your case, if XLATE_TOO_MANY_RESUBMITS happens, the 'drop' flow will be installed into kernel datapath and the actual packet drops will happen in the kernel. Coverage counters can't count these. So, drop_action_* counters are only useful for userspace datapath at the moment. Best rgards, Ilya Maximets. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
