IPSK Manager and High Availability

[ds_livesay]

Hello Everyone,

I am currently working on a IPSK Manager deployment and want to setup some type of HA for the system.  I have a primary server successfully working for self registration.  I want to setup a second IPSK Manager server for HA purposes and have all users, portals ( I know I might need to build the portals manually) and MAC addresses replicated over to the HA server in case of a disaster.  However I want to see about getting two way replication working so that when the primary server is recovered everything will replicate back to the primary server and it can take over.  The GitHub page for IPSK Manager references briefly that HA is possible with the built in features of MYSQL.  However, I wanted to see if anyone has tried this? Any catches?  What type of MySQL replication did you use?  Is there a better way?  I have been looking into Master - Master MySQL replication.  If anyone has MySQL in depth knowledge, is this what I should use?  Thank you for any feedback.  It will help with assessing my options before trying to deploy IPSK Manager into production.

[Arne Bier · Accepted solution]

I am going back some years now, but there are two things in play:

1. SQL Database replication occurs between servers SQL1 and SQL2 (using the example of two servers) - I think this is built-in to mySQL these days. I might be wrong.
2. Your Applications will be communicating to your SQL "cluster" using a virtual IP (i.e. your applications are unaware that there are two SQL servers) - this requires more work.  See below

For MySQL most folks ended up creating a homebrew solution using Pacemaker and Corosync. The reason this is a homebrew is because in the simplest case there is no centralised storage (e.g. a fiber channel attached storage volume). If you don't have a centralised storage solution for servers SQL1 and SQL2, then you have to replicate tables between them. And Pacemaker and Corosync are used to create a VIP (Virtual IP) that creates the illusion of a cluster. Your application talks to the VIP and doesn't need to care which SQL server is the active one.

It's probably out of scope for this forum but I think there are some more up to date examples on the net.

[ds_livesay · Accepted solution]

Thank you for the response. That sounds like a good solution if I want a separate sql server. In this case the sql server and application server are the same. I just want replication for DR purposes. What built in option are there with mysql? I was trying to avoid having to use a third-part app to do this.

[Arne Bier · Accepted solution]

The topic of MySQL replication seems to be covered by experts out there and it can be done without any third party ($$$) tools. 
see if this link helps?

https://www.toptal.com/mysql/mysql-master-slave-replication-tutorial