Guys, this would work, yes, as Igor also mentioned, the only thing I'm concerned (but will need to test in on dev enviroment) is if NFS share that is already mounted on KVM hosts, will be gratefully remounted and not stuck/stale when I move IP from one NFS box to another...
Anyone has experience with this ? Thanks, Andrija On 4 February 2016 at 06:00, Sanjeev N <sanj...@apache.org> wrote: > Hi Daniel, > > It is possible to have the MAC-IP pair in the switch arp cache. You can use > arping from your new NFS server to send gratious arp request, so that the > switch can update the MAC address in its arp cache. > > Thanks, > Sanjeev > > On Thu, Feb 4, 2016 at 1:47 PM, Daniel Mezentsev <d...@soleks.com> wrote: > > > Igor, > > > > I have question. Doing such IP migration is it possible that arp cache on > > the switch (assuming that we have layer 3 switch) can have old server > pair > > MAC<--->IP. > > > > > > Andrija, as for me the best solution: > >> > >> 1. shutdown data interfaces (or bond) on switch for new server after > >> > > rsync > > > >> 2. Configure same IP as for old server > >> 3. via switch CLI shutdown data ports for old server > >> 4. via switch CLI startup ports for new server > >> 4. Reconfigure IP for data bond old server via mgm interface to new IP > >> 5. via switch CLI startup data ports for old server on switch > >> > >> All this can be done via ansible. Downtime will be only between steps 3 > >> and > >> 4. And it's will be around 0,1-0,5 seconds. > >> > >> From cloudstack mgmt sever it's will be transparent. Possible 1-2 TCP > >> retransit will happen. > >> > >> 2016-02-03 22:49 GMT+02:00 Andrija Panic <andrija.pa...@gmail.com>: > >> > >> Hi guys, > >>> > >>> I need to do manitance of 1 Seconary Storage NFS server fro few days, > >>> so I > >>> though to temporarily rsync data to another NFS box, and switch IP > >>> addresses, so the new NFS box has the original IP for few days...(need > >>> > >> to > > > >> test if KVm nodes will gracefullt remount the NFS server during the IP > >>> switchover...) > >>> > >>> Is this the prefered way, or should I hack the DB to point existing NFS > >>> server (that is defined in ACS) to the new IP of the second NFS box, > and > >>> perhaps restart mgmt and agents across all nodes. > >>> > >>> Any recomendations ? > >>> > >>> Best, > >>> > >>> Andrija Panić > >>> > >> > >> > >> > > Daniel Mezentsev, founder > > (+1) 604 313 8592. > > Soleks Data Group. > > Shaping the clouds. > > > -- Andrija Panić