Research: Failover support within the current IPS implementation
Failover support would make sure that if there are multiple Suricata instances and one of those disappears, the other Suricata IPS instances would 'do the right thing'. What doing the right thing is is an open question. Some of the ideas:
- sync flow table so that flow tracking would stay active
- sync 'drop settings' per flow/host/etc
- sync thresholding
I think one of the first things that needs to be done is analyze how Suricata currently works in a IPS failover case.
Updated by Jason Ish almost 2 years ago
Existing fail over solutions may be worth looking into.
CARP is for failover of IP addresses for firewalls. It uses a virtual IP address for the firewall, then when one fails, the spare takes over that IP address. This might be re-useable as is in a IPS setup, especially if routing is used.
But CARP is only for the address redundancy. PF has a tool called "pfsync" (see https://www.openbsd.org/faq/pf/carp.html) that takes care of syncing the firewall state. This is probably a close example to what would be needed between 2 Suricata instances.
I'd also start by the bare minimum, perhaps stuff like datasets could be in a phase 2 on the understanding that failover may not sync all state.