A simple way to do the DR?

[Jim2015]

Hi,

I am finding a simple way to do the DR, assuming there is a solution like below:
1. Setup two zabbix server, one in product environment, one in backup environment, there is a firewall between two environments, the firewall deny the communication between agent and the backup zabbix server in normal time.
2. Configure both 2 server IP in agent.
3. Do the backup for product zabbix regularly.
4. when product zabbix breaks down, restore the backup data to backup zabbix
5. Open the ports on firewall.
6. the zabbix service comes back.
7. when product zabbix is repaired, close the firewall, back/restore data from backup zabbix to product zabbix.

The solution is easy to implement(the cost is increase RTO/RPO), but there are some questions:
1. As one of the servers is disconnected at anytime, the agent will always retry in the 'timeout' period. Then will it increase the agent process load and effect the host's performance? even effect the agent normal functions?
2. The firewall will be forced to handle a large number of requests, will it be under attack like DDOS ?

Considering these problems, it seems that use FQDN in ServerActive is a better way? When the product zabbix breaks down, only to change the IP from product server to backup server in DNS server.

- Jim

[Jim2015]

Hi Experts,

I'd like to have your suggestion, is it a possible solution? I very appreciate your help if you post your advises here!

- Jim

[Jim2015]

I hope what I have posted is possible

[ingus.vilnis]

Hi Jim,

Your given solution is doable but impractical and has lots of drawbacks. I would not recommend you such setup.

You could build a HA cluster instead.
Here are some pointers to do so:

http://zabbix.org/wiki/Docs/howto/high_availability

or

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/pdf/Cluster_Administration/Red_Hat_Enterprise_Linux-6-Cluster_Administration-en-US.pdf

The idea very shortly:

• Install two servers with Zabbix
• Use shared database on a mapped location
• Install cluster resource management software
• Configure Virtual IP to be used as server IP for agents to connect
• Create a DB clustering/backup separately
• Configure the Zabbix cluster nodes to failover automatically in case any service or connectivity is lost
• Configure Zabbix agents to connect to one Virtual IP. Which real IP is used will be determined by cluster management tools.

FQDN can be used but for each connection to server the agent will need an extra hop to DNS server which takes time, sometimes is not reolving and think, what will happen if the DNS server goes down completely? Better use IP instead.

Hope this helps!

Best Regards,
Ingus