Elastix HA Cluster
-
Something to consider is if you do an HA cluster of this nature, that makes this "all in one place" and you can't fail over between sites. That's why we rarely do this. That and it is a stateless system so we normally just use the platform for it as that is less complex.
-
@scottalanmiller yes, as you mentioned Scott .. but what are the considerations we should have if we need to setup fail over clustering between sites?.
-
@AlyRagab said in Elastix HA Cluster:
@scottalanmiller yes, as you mentioned Scott .. but what are the considerations we should have if we need to setup fail over clustering between sites?.
You can't use thinks like DRBD. You use Rsync and you generally fail over manually to ensure no split brain. But it only takes seconds and you CAN automate it with a script instead.
-
@scottalanmiller said in Elastix HA Cluster:
@AlyRagab said in Elastix HA Cluster:
@scottalanmiller yes, as you mentioned Scott .. but what are the considerations we should have if we need to setup fail over clustering between sites?.
You can't use thinks like DRBD. You use Rsync and you generally fail over manually to ensure no split brain. But it only takes seconds and you CAN automate it with a script instead.
I do something similar with my KVM hosts. They are running Gluster instead of Rsync. But if one were to die I just manually start it on the other with
ansible host -m shell -a "virsh start vm_name" --become --ask-become-pass -
@johnhooks said in Elastix HA Cluster:
@scottalanmiller said in Elastix HA Cluster:
@AlyRagab said in Elastix HA Cluster:
@scottalanmiller yes, as you mentioned Scott .. but what are the considerations we should have if we need to setup fail over clustering between sites?.
You can't use thinks like DRBD. You use Rsync and you generally fail over manually to ensure no split brain. But it only takes seconds and you CAN automate it with a script instead.
I do something similar with my KVM hosts. They are running Gluster instead of Rsync. But if one were to die I just manually start it on the other with
ansible host -m shell -a "virsh start vm_name" --become --ask-become-passYou are running Gluster over WAN links?
-
With Elastix HA you would normally have all VMs up all of the time, synced and then only start the Asterisk service for failover, not the VM itself.
-
@scottalanmiller said in Elastix HA Cluster:
@johnhooks said in Elastix HA Cluster:
@scottalanmiller said in Elastix HA Cluster:
@AlyRagab said in Elastix HA Cluster:
@scottalanmiller yes, as you mentioned Scott .. but what are the considerations we should have if we need to setup fail over clustering between sites?.
You can't use thinks like DRBD. You use Rsync and you generally fail over manually to ensure no split brain. But it only takes seconds and you CAN automate it with a script instead.
I do something similar with my KVM hosts. They are running Gluster instead of Rsync. But if one were to die I just manually start it on the other with
ansible host -m shell -a "virsh start vm_name" --become --ask-become-passYou are running Gluster over WAN links?
Oh no, I just meant I'm not syncing with DRBD and manually starting at the other end.
-
@scottalanmiller Using Rsync will make us stop the mysql daemon that will cause a downtime, and in case the master server went down how we make the clients connect to the second server?.
-
@AlyRagab said in Elastix HA Cluster:
@scottalanmiller Using Rsync will make us stop the mysql daemon that will cause a downtime, and in case the master server went down how we make the clients connect to the second server?.
You wouldn't rsync the mysql database, you use the tools built into mysql for replication.
-
@AlyRagab said in Elastix HA Cluster:
@scottalanmiller Using Rsync will make us stop the mysql daemon that will cause a downtime, and in case the master server went down how we make the clients connect to the second server?.
Clients can be connected to two at once. Or you use DNS to switch. Either way works just fine.
Rsync does need MySQL to stop IF you are copying that data all of the time. But do you need to? We don't. We only need VoiceMail to be synced (and only sometimes) and you only need MySQL synced on changes, not on normal usage. So only occasional MySQL stops that can be combined with updates and reboots under normal circumstances.
