|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
X-Original-To: nanog@nanog.org From: David Bass <davidbass570@gmail.com> In-Reply-To: <56C5E869.7030406@foobar.org> Date: Thu, 18 Feb 2016 11:19:02 -0500 To: Nick Hilliard <nick@foobar.org> Cc: NANOG <nanog@nanog.org> Errors-To: nanog-bounces@nanog.org I don't think I'd trust any vendor's "ISSU" to be completely without impact.= ..been more of a marketing term from my experience... > On Feb 18, 2016, at 10:51 AM, Nick Hilliard <nick@foobar.org> wrote: >=20 > Jason Bothe wrote: >> The 9k does however get a huge win with the ability to apply a =E2=80=98p= ie=E2=80=99 >> or software patch while staying in service vs requiring a reload. >=20 > SMUs are often "hitless", which is to say, "hitless" with scary quotes. > What this means in practice is that the SMU itself might be hitless but > it will depend on 47 other SMUs, thereby almost guaranteeing some form > of reload. Also, restarting processes is "hitless" (e.g. restarting > bgpd, ospfd, etc) or shutting down interfaces. >=20 > E.g.: >=20 > CSCuo47663: "Hitless/Optional SMU,aigp metric different in RIB & BGP > table". This will restart the bgp process. >=20 > CSCus26923: "traffic from SIP700 to 9000v is dropped when a link to > 9000v flaps". Release notes state that the issue is not service > impacting, then "After the SMU installation , we need to apply > shut/noshut of the problematic interface to trigger the hardware > programming." Wuh?? >=20 > In other words, "hitless" does not mean "not service impacting". >=20 > Nick
|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |