HA ('ish) enviroment

[baardman]

Hello!

We are running a hosting company on a CentOS vm with Directadmin.

We currently switched from one VM to the following server setup:

Node1: 4 cores 8 threads 32 gb ram (disaster'ish system) (power A+B feed)
4x 100 GB Consumer SSD

Node2: 16 Cores 32 Threads 255 GB Ram (Main production fast server) (power A Feed)
4x 4TB Server HDD (WD GOLD)

Node3: 16 Cores 32 Threads 255 GB Ram (Possible fast server with maybe fallback potential or something?) (power B feed)
4x 4TB Server HDD (WD GOLD)

NodeStorage: 6 cores 12 threads, 150 GB Ram (Storage server) (Power A + B feed)
8x 2TB Datacenter grade SSD, Kingston DC
Alot of 4TB HDD for backup potential.

We also have redundant Switches setup by NOC specialists.

We currently share the NodeStorage over NFS to the other VM's for fallback potential.

What is the best way to make sure the VM that runs on Node2 doesn't `ever`(or the most posible) shuts down in the course of an event.

We are looking at the options of ZFS share but are running into a couple of problems, since the disk is named vm 100, and if node 2 shuts down, we need to delete the vm and migrate it. We want to automate this as best as possible, so there are no PEBCAKS.

Please help us.
Thanks!

 

[baardman]

BUMP,

Someone can help me with this?

 

[tom]

You cannot expect that the forum community designs your hosting setup.

two comments:

• You run consumer storage on your nodes, (SSD, HDD), not suited if you expect maximum reliability and performance
• Your storage box is a single point of failure, consider HA storage like Ceph.

Before configuring HA, I recommend to use high quality hardware.

 

[baardman]

I understand, but i just want another look on it.

What makes Ceph a better alternative?

i plan on cloning the vm to the HDD’s in the blade servers, and when the storage vault is down, we still run on the HDD’s.

ofcourse i understand the consumer ssd’s are not for serving tasks, we plan to run a windows VM for migration purposes so we enjoy the fast internet