08-11-2015 02:34 AM
Is there a document somewhere which explains Master failover (using VRRP) in more detail? In particular I'm wondering about the effect on Locals and if there should be any noticeable interruption to service.
The reason I ask is we have a pair 3600s as Master and Standby Master using VRRP with preemption. We then have 5x 7220 Locals (using VRRP to fail over the LMS, but that's not important). The Masters do not handle any APs - just the configuration of the Mobility Domain.
When I disconnected the Master, the Locals took a little time to reestablish communications with the new Master (the Standby Master that takes over). There seemed to be a longer delay when the original Master came back online (after I moved it to a new location in the data centre).
1. How quickly should the Locals reestablish communications with the new Master? The Locals were showing 'Master Unreachable' for a minute or so afterwards (when VRRP on the new Master had failed).
2. When the Locals reestablished communications following a Master change, they appeated to download the configuration from it, even though it was the same across the Master and Standby Master before failover occurred (it had been save, waited for it to synchronise and the Config ID was the same). Is this to be expected?
3. When the original Master comes back online, it presumably has to check with the Standby Master to see if there have been any configuration changes whilst it's been down and synchronise them back?
4. If this is the case, will there by any problems with the returning Master taking over the VRRP address before this process is complete? If so, is this a reason to use 'preempt delay ...' (to give the returning Master time to synchronise up and be ready to talk to the Locals, before it takes over the Master role)?
5. Should there be any actual service outages during this process? In general, I stayed connected without issue, but there was a point where I got kicked off an 802.1X SSID (but not immediately). It was difficult to tell what was actually going on across all the controllers at this time as I didn't have network access from the point where I was.
Thanks in advance, and for any further information you can give.
08-11-2015 03:31 AM
First of all, Master-Standby failover will not effect the local operation, there will not be any outage.
1. Maser failover delay depends on LAN (VRRP), as a best practice we need to disable the preemption.
2. Database synchronisation will happen whenever we run "write memory" command on the master (Primary or Standby) or by default for every 60 mins and which is configurable.
3. If the database is synchronised they should have the same config hence, for a local it doesn't matter whether it is communicating with master or standby.
Hope you got some more clarity on this, please feel free for nay further query on this.
[Is my post helped you ? Give Kudos :) ]