It's immediate in most cases.

Imagine this setup:

pf1: mgmt ip 192.168.0.1

pf2: mgmt ip 192.168.0.2

mgmt vip: 192.168.0.3

Keepalived manage this vip

now on both servers you have radius-auth (192.168.0.1,192.168.0.2 port 
1812), radius-acct (192.168.0.1,192.168.0.2 port 1813) and 
radius-load-balancer (192.168.0.3 port 1813 and port 1812) that proxy 
radius request to (192.168.0.1,192.168.0.2 port 1812 and port 1813).

So on each you have radius-load-balancer listening on ip 192.168.0.3 , 
so depending where is the vip one of them proxy radius request to 
himself and the other one.

We also use monit to check if the radius load balancer is still running 
(https://github.com/inverse-inc/packetfence/blob/devel/addons/monit/30_packetfence-activeactive.sh).


So the in the better case if one of the server crash then the vip will 
move to the other one and radius-load-balancer will start to process and 
if the other server is back again then radius-load-balancer will start 
to forward radius request to it.

If the radius-load-balancer crash where is the vip then monit will save you.

Btw internally to radius-load-balancer there is some check to see if 
radius-auth and radius-acct are alive and stop/start proxying radius 
request depending of the status.

Regards

Fabrice



Le 2016-10-12 à 17:10, Sallee, Jake a écrit :
>> Btw in active setup if one of the server crash (not just radius) then
> the second server will fetch the vip and start to process radius on the
> radius load balancer process.
>
> Does that mean the cluster has no way to know if only a single service fails?
>
> For example:  If my radiusd service stops but the other services are still 
> up, will the other server then start answering all radius requests?
>
> Or does it take a complete failure before the other server takes over?
>
> Jake Sallee
> Godfather of Bandwidth
> System Engineer
> University of Mary Hardin-Baylor
> WWW.UMHB.EDU
>
> 900 College St.
> Belton, Texas
> 76513
>
> Fone: 254-295-4658
> Phax: 254-295-4221
>
> ________________________________________
> From: Durand fabrice <fdur...@inverse.ca>
> Sent: Tuesday, October 11, 2016 5:25 PM
> To: packetfence-users@lists.sourceforge.net
> Subject: Re: [PacketFence-users] Active/Active PF cluster with Maria Galera
>
> Hello Jake,
>
> i am probably the only one at inverse who did a MariaDB Gallera cluster
> with PacketFence.
>
> In this setup i have 4 servers, 2 for PacketFence (garb is intalled on
> one of them) and 2 for the MariaDB gallera cluster and the idea was to
> load balance with haproxy the sql connections between both db server.
> (PacketFence sql request vs radius sql request).
>
> So your assumptions are right, most of PacketFence's services are
> running active/active (except pfmon and pfdhcplistener) and the db only
> run on one of them (2 servers setup).
>
> Btw in active setup if one of the server crash (not just radius) then
> the second server will fetch the vip and start to process radius on the
> radius load balancer process. (you have 3 radius process, one for auth,
> another one for acct and the last one for load balancing).
>
> The radius load balancing process will act as a proxy to load balance
> the radius request on both server, haproxy will do the same for
> http/https request.
>
> If you need more details fell free to ask more questions.
>
>
> Regards
>
> Fabrice
>
>
>
>
>
>
>
> Le 2016-10-11 à 17:46, Sallee, Jake a écrit :
>> I am reading through the Active/Active documentation and it mentions that 
>> the DB portion can be made active/active using Maria Galera.
>>
>> What are the benefits/drawbacks of going with the MariaDB route vs the MySQL 
>> route?
>>
>> I read in the MariaDB Galera docs that you really need at least 3 servers.
>>
>> Also, if I am reading the info correctly, the clustering is setup so all the 
>> services other than the DB are running in an active/active capacity, right?
>>
>> So if one of my servers has the radius process stop for whatever reason the 
>> other will keep going and while they are both running they can share the 
>> load correct?
>>
>> Sorry for the scattered nature of this message, I'm pressed for time ATM.
>>
>> Please feel free to chime in with any and all feelings and advice ... on 
>> this issue.
>>
>> Jake Sallee
>> Godfather of Bandwidth
>> System Engineer
>> University of Mary Hardin-Baylor
>> WWW.UMHB.EDU
>>
>> 900 College St.
>> Belton, Texas
>> 76513
>>
>> Fone: 254-295-4658
>> Phax: 254-295-4221
>>
>> ------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, SlashDot.org! http://sdm.link/slashdot
>> _______________________________________________
>> PacketFence-users mailing list
>> PacketFence-users@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/packetfence-users
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, SlashDot.org! http://sdm.link/slashdot
> _______________________________________________
> PacketFence-users mailing list
> PacketFence-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/packetfence-users
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, SlashDot.org! http://sdm.link/slashdot
> _______________________________________________
> PacketFence-users mailing list
> PacketFence-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/packetfence-users


------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most 
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
PacketFence-users mailing list
PacketFence-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/packetfence-users

Reply via email to