Re: [Astlinux-users] Using WAN Failover for Astlinux High Availability

[Michael K. <mic...@ip...>]

Yes that could be used too although it would require lots more testing.

Regards
Michael Knill

From: Michael Keuter <li...@mk...>
Reply-To: AstLinux List <ast...@li...>
Date: Monday, 2 July 2018 at 8:51 am
To: AstLinux List <ast...@li...>
Subject: Re: [Astlinux-users] Using WAN Failover for Astlinux High Availability



Am 02.07.2018 um 00:12 schrieb Michael Knill <mic...@ip...<mailto:mic...@ip...>>:

Hi Michael

Sorry I probably shouldn't have called it HA as that is more as you described. Maybe it would be better calling it System Failover instead.
Keeping them in sync should be pretty easy and I can just exclude files that are different. I would make the failover time fairly long so it only did it when it was truly broken. The failover script would enable the interfaces which are addressed the same as the primary. Management would be via a separate interface.

So do you think it could work with WAN Failover?

Needs to be tested very thoroughly.
Maybe the clients will cache not only the IP-address of the server, but also the MAC-address.

Another approach would could be the "Server 1" + "Server 2" option, that many IP-phones offer (never tested it though), to get around the IP-address problematic:

[cid:AF8...@pr...]


Regards
Michael Knill

On 1/7/18, 9:07 pm, "Michael Keuter" <li...@mk...<mailto:li...@mk...>> wrote:



Am 01.07.2018 um 01:42 schrieb Michael Knill <mic...@ip...<mailto:mic...@ip...>>:

Hi Group

Tell me if I am totally off track but I have been wondering for a while how I can do Astlinux HA e.g. having a primary and standby box
So what do you need for this:
• Monitoring of the primary Astlinux server
• Run a script if connectivity lost
• Some sort of timers to prevent flapping
• Email notification to indicate that it has occurred
• Config sync between the two

When I considered the list, isn’t 1) to 4) actually already available with WAN Failover? Your monitor IP is the address of the other box?
Can you failover to the same interface?

Interesting thought!

Regards
Michael Knill

   Hi Michael,

   for "real" HA clusters usually a "Virtual IP" is used (where the SIP clients connect to), which is then "switched" to to real IP of the actice node.
   For Linux there are a few solutions:

   https://en.wikipedia.org/wiki/Common_Address_Redundancy_Protocol
   https://github.com/jedisct1/UCarp  (though the "ucarp.org<http://ucarp.org>" domain seem to now go to some Asian site)
   https://serverfault.com/questions/276170/alternatives-to-heartbeat-pacemaker-and-corosync

   I also got requests from a few customers a while ago, and we found it simpler just to use 2 identical boxes configure them the equally and put one offline in the shelf.
   In case of problem just switch the boxes.
   The only issue is to keep the boxes in sync (because both have the same IP).
   We used rsync to a server share, switched of the actice box for a few minutes afterhours and occacionally synced the offline box to that (we excluded a few hardware/MAC related files).
   You also can give them 2 different IPs, and change the IP in case of a problem.

   But to be honest: I do Asterisk/AstLinux stuff since 2007, but except an old water damage case (with only one net5501 under water that survived, but not the internal ISDN PCI card), I never had an issue that an AstLinux box broke and we really needed the Failover, except for testing, that it actually works as expected :-).
   In all other problem cases either the power was down, the ISP/internet connection was down/had issues or the SIP provider had problems :-).

   Michael

Michael

http://www.mksolutions.info