[cisco-voip] Expressway Cluster failover for MRA...

Wed Jan 29 13:10:55 EST 2020

Correct, the edge cluster and the control cluster are very much separate entities apart from themselves and load balance within their own respective clusters.

As long as clustering has been implemented correctly, in that scenario, the edge cluster would still be able to load balance and distribute resources within its own cluster (edge 1 answered the request but edge two transported the request to control 2 because control 1 was unreachable).

In that type of a degraded “brown out” scenario, you should manually put edge 1 in maintenance mode. Not sure if you could trigger maintenance mode via API but if you could, then its conceivable that one could write a polling “heart beat” script to detect those unique “brown out” states and then automatically make edge 1 unserviceable by placing it into maintenance mode.

Thanks,

Ryan

On Jan 29, 2020, at 12:59, ROZA, Ariel <Ariel.ROZA at la.logicalis.com> wrote:

But without clustering, if Core1 fails, Edge1 will still be active and Jabber clients will still see Edge1 running and attempt to connect through it!

De: cisco-voip <cisco-voip-bounces at puck.nether.net> En nombre de Charles Goldsmith
Enviado el: martes, 28 de enero de 2020 23:18
Para: Lelio Fulgenzi <lelio at uoguelph.ca>
CC: cisco-voip at puck.nether.net
Asunto: Re: [cisco-voip] Expressway Cluster failover for MRA...

We've built them as individual pairs (Edge/Core) and then use DNS to control which one goes where.  Without the cluster, we know that Edge1 will always talk to Core1.

I get the feeling that clustering was always meant to be in the same DC, and for redundancy purposes in the same DC.

If you have two DC's, either a cluster at each DC, or just a pair at each DC, depending on the business needs.

On Tue, Jan 28, 2020 at 8:11 PM Lelio Fulgenzi <lelio at uoguelph.ca<mailto:lelio at uoguelph.ca>> wrote:

How does no. 2 actually solve the problem of having to log back in?

Is this a supported/suggested deployment method?

It’s been a while since I first looked at things and don’t recall things mentioning using the cluster name in the SRV records.

I’m intrigued. And interested!

-sent from mobile device-

Lelio Fulgenzi, B.A. | Senior Analyst
Computing and Communications Services | University of Guelph
Room 037 Animal Science & Nutrition Bldg | 50 Stone Rd E | Guelph, ON | N1G 2W1
519-824-4120 Ext. 56354<tel:519-824-4120;56354> | lelio at uoguelph.ca<mailto:lelio at uoguelph.ca>

www.uoguelph.ca/ccs<https://nam02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.uoguelph.ca%2Fccs&data=02%7C01%7C%7C28abee85602c4394504a08d7a4e5024d%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637159175814864913&sdata=4Ty7ZYZ77QBJ6FCaao%2FQOZRlnNj7Oi4y%2BOBQzilxEIA%3D&reserved=0> | @UofGCCS on Instagram, Twitter and Facebook

On Jan 28, 2020, at 9:03 PM, Ryan Huff <ryanhuff at outlook.com<mailto:ryanhuff at outlook.com>> wrote:
1.) It used to be in previous versions that all cluster nodes could technically be active at any time and SRV weights and priorities could influence the path selection but not guarantee it end-to-end when all cluster nodes are up and running.

I believe this behavior has changed/improved and I think you are supposed to be able to control that now with SRV weights and priorities, but I could be wrong. I haven’t played with Expressway clustering in a bit.

2.) As far as the Jabber registration goes; what I’ve done before in the edge is have the collab-edge SRV point to the edge cluster FQDN as the target. Then I create round robin A records for the cluster FQDN (one resolving your each edge server). The for the edge certs, just make sure the edge cluster fqdn is in the SAN.

This way if one of the edge server goes down, the Jabber client is ultimately still trying to resolve the same MRA FQDN via SRV lookup (this a key to Jabber client failover for MRA).

Thanks,

Ryan

On Jan 28, 2020, at 20:50, Jonathan Charles <jonvoip at gmail.com<mailto:jonvoip at gmail.com>> wrote:

We have two pairs of Expressway clusters (C/E) at two different locations (primary and DR)...

The cluster is up, however, we want to make sure that we are in Active/Standby.

Currently, we have one of our SRV records for collab-edge set at 5 (the backup is at 10) with the same weight.

The clustering guide says we should set the priority and weight on both SRV records the same, which will cause half of the registrations to go to the DR site. It is far away and has less capability.

How do we:

1 - Make sure the primary site handles all MRA registrations and the DR site is only used when the primary is down.
2 = Make sure failover occurs automatically... currently Jabber users have to log out and back in to connect to the DR site.

Thanks!

Jonathan

_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpuck.nether.net%2Fmailman%2Flistinfo%2Fcisco-voip&data=02%7C01%7C%7C2f536d8162984707853908d7a45d8e24%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637158594035084563&sdata=atRtIR8sWZ60Ja8akD6GjzBIgBNC8GSJjaOmu%2BTxmWw%3D&reserved=0<https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpuck.nether.net%2Fmailman%2Flistinfo%2Fcisco-voip&data=02%7C01%7C%7C28abee85602c4394504a08d7a4e5024d%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637159175814874918&sdata=2ULQiab5aIZYA90C1VDgpALdcRZjlsP1Vwx%2FCAam4Qs%3D&reserved=0>
_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://puck.nether.net/mailman/listinfo/cisco-voip<https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpuck.nether.net%2Fmailman%2Flistinfo%2Fcisco-voip&data=02%7C01%7C%7C28abee85602c4394504a08d7a4e5024d%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637159175814885738&sdata=IjY%2FpOIyx%2FZzOTxUuruUDEx9UHH%2Bl0QvfLsf8K2nqQI%3D&reserved=0>
_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net<mailto:cisco-voip at puck.nether.net>
https://puck.nether.net/mailman/listinfo/cisco-voip<https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpuck.nether.net%2Fmailman%2Flistinfo%2Fcisco-voip&data=02%7C01%7C%7C28abee85602c4394504a08d7a4e5024d%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637159175814894934&sdata=VNEu5UEe0dooSVh63Wbxx4xQope8qS0oV5xlDOZat%2FA%3D&reserved=0>
_______________________________________________
cisco-voip mailing list
cisco-voip at puck.nether.net
https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpuck.nether.net%2Fmailman%2Flistinfo%2Fcisco-voip&data=02%7C01%7C%7C28abee85602c4394504a08d7a4e5024d%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637159175814944977&sdata=8geL2Lp2ksdsbzZ0lppv5z6Eub6QbdAGjiAjtsTshNY%3D&reserved=0
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/cisco-voip/attachments/20200129/165f4e97/attachment.htm>