ericbrunson
Technical User
I used to be an old pro at SunOS back in the 4.1.3 days, but I'm tooling up on Solaris 10 now and I'd like input on some of the HA features available in Sun Cluster.
I have a (very) high transaction MySQL database running in a Solaris 10 zone on one physical server. I'm using MySQL replication to mirror the data over to another identical zone on a different server. I have a virtual IP address that is plumbed on the host of the mysql master and all database connections use that VIP.
On the original Linux architecture we were using heartbeat and some custom monitoring scripts to evaluate the health of the mysql master. When the master failed we would use IP takeover to move the database VIP over to the slave and run in a degraded mode until the master could be restored, at which time the master would actually become the slave and the hot standby.
Our zones were configured by someone else in the organization, and I'll admit I only have a cursory understanding of the network plumbing between my zone and the global zone. However, it seems that in order to move the VIP between machines it requires a bounce of the (non-global) zones. If I'm incorrect on that assumption, please let me know.
I've been told that the Sun Cluster package would allow us to transfer that VIP back and forth freely. And I do mean freely, because I want the freedom to change the role of master and slave without interactive access to the global zones. Ideally, scripts I've written (running in the global zone if required) would be adapted to notice if I take down the master and fail the VIP over to the standby node. I'll deal with all the intricacies of the mysql, but I need to be able to induce the failover based on the availability of the mysql master instance.
My question is pretty simple. Is this feasible or even possible? Most of the documentation I've read about clustering with zones involves putting the zone on shared storage and having the global zones share it. The I/O required by the database precludes anything but local disk access, and because of my familiarity with mysql replication, I'm simply more comfortable with the application level redundancy I've implemented in the past.
I'm open to other suggestions, but I think this solution would best fit our needs. Can anyone outline the process to set this up and/or refer me to any documentation covering it?
Thanks for your time.
I have a (very) high transaction MySQL database running in a Solaris 10 zone on one physical server. I'm using MySQL replication to mirror the data over to another identical zone on a different server. I have a virtual IP address that is plumbed on the host of the mysql master and all database connections use that VIP.
On the original Linux architecture we were using heartbeat and some custom monitoring scripts to evaluate the health of the mysql master. When the master failed we would use IP takeover to move the database VIP over to the slave and run in a degraded mode until the master could be restored, at which time the master would actually become the slave and the hot standby.
Our zones were configured by someone else in the organization, and I'll admit I only have a cursory understanding of the network plumbing between my zone and the global zone. However, it seems that in order to move the VIP between machines it requires a bounce of the (non-global) zones. If I'm incorrect on that assumption, please let me know.
I've been told that the Sun Cluster package would allow us to transfer that VIP back and forth freely. And I do mean freely, because I want the freedom to change the role of master and slave without interactive access to the global zones. Ideally, scripts I've written (running in the global zone if required) would be adapted to notice if I take down the master and fail the VIP over to the standby node. I'll deal with all the intricacies of the mysql, but I need to be able to induce the failover based on the availability of the mysql master instance.
My question is pretty simple. Is this feasible or even possible? Most of the documentation I've read about clustering with zones involves putting the zone on shared storage and having the global zones share it. The I/O required by the database precludes anything but local disk access, and because of my familiarity with mysql replication, I'm simply more comfortable with the application level redundancy I've implemented in the past.
I'm open to other suggestions, but I think this solution would best fit our needs. Can anyone outline the process to set this up and/or refer me to any documentation covering it?
Thanks for your time.