Ben Pfaff <b...@ovn.org> wrote on 07/27/2016 12:30:25 PM:
> From: Ben Pfaff <b...@ovn.org> > To: Hui Kang/Watson/IBM@IBMUS > Cc: Hui Kang <hkang.sun...@gmail.com>, dev@openvswitch.org > Date: 07/27/2016 12:30 PM > Subject: Re: [ovs-dev] [PATCH v3] Scanning only changed entries in the ovnsb > > On Tue, Jul 26, 2016 at 03:44:57PM -0400, Hui Kang wrote: > > > > > > "dev" <dev-boun...@openvswitch.org> wrote on 07/26/2016 02:20:27 PM: > > > > > From: Ben Pfaff <b...@ovn.org> > > > To: Hui Kang <hkang.sun...@gmail.com> > > > Cc: dev@openvswitch.org > > > Date: 07/26/2016 02:20 PM > > > Subject: Re: [ovs-dev] [PATCH v3] Scanning only changed entries in the > > ovnsb > > > Sent by: "dev" <dev-boun...@openvswitch.org> > > > > > > On Sat, Jul 16, 2016 at 11:58:25PM -0400, Hui Kang wrote: > > > > Improve performance by scanning only changed port binding entries > > > > when determining whether to mark the logical switch port up or > > > > down > > > > > > > > Signed-off-by: Hui Kang <ka...@us.ibm.com> > > > > > > Won't this skip an initial round of updates at ovn-northd startup time? > > > (Certainly ovn-northd might get killed and restarted occasionally, > > > especially if we're doing failover to a second host.) > > > > Hi, Ben, > > After second thought, I think skipping the initial round is the purpose of > > this patch. > > > > ovsdb_idl_create(ovsdb) copies the the Port_binding table from southbound > > database whenever ovn-northd gets started. In this case, the northbound > > DB and southbound db are synced. > > I don't understand this statement. ovsdb_idl_create() doesn't sync > anything between nb and sb dbs. Sorry for the confusion. I have updated the patch to v4 addressing your comment. - Hui > _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
