Storage migration from Raft Integrated Storage to PostgreSQL

Vault
1

Hi

We encountered an issue which causes significant production downtime since a network error causing the raft replication to failed, even though we are running on 6 nodes. We are looking for an alternative storage backend that would reduce this issue. I stumbled upon PostgreSQL backend and became interested in it.

I’m running a cluster consisted of a bunch of 1.19.x. Here is the result of vault operator members:

Host Name       API Address                   Cluster Address               Active Node    Version    Upgrade Version    Redundancy Zone    Last Echo
---------       -----------                   ---------------               -----------    -------    ---------------    ---------------    ---------
vault-node6     https://172.16.xx.xx:8200     https://172.16.xx.xx:8201     false          1.19.3     1.19.3             n/a                2025-05-15T14:22:09+07:00
vault-node1     https://172.16.xx.xx:8200     https://172.16.xx.xx:8201     true           1.19.0     1.19.0             n/a                n/a
vault-node5     https://172.16.xx.xx:8200     https://172.16.xx.xx:8201     false          1.19.3     1.19.3             n/a                2025-05-15T14:22:08+07:00
vault-node3     https://172.16.xx.xx:8200     https://172.16.xx.xx:8201     false          1.19.3     1.19.3             n/a                2025-05-15T14:22:10+07:00
vault-node4     https://172.31.xx.xx:8200     https://172.31.xx.xx:8201     false          1.19.1     1.19.1             n/a                2025-05-15T14:22:06+07:00
vault-node2     https://172.31.xx.xx:8200     https://172.31.xx.xx:8201     false          1.19.3     1.19.3             n/a                2025-05-15T14:22:06+07:00

I’ve seen some other posts mentioning that PostgreSQL is not recommended since the storage layout is “directories”, therefore the integrated Raft storage should be a much better option.

I’d also like to mention that my Raft database’s size (I suppose the BoltDB database) is around 16 GB.

Questions:

  1. Is there any migration path that can be done from the Raft Integrated storage to PostgreSQL storage?
  2. If migrating to PostgreSQL is not recommended and it’s recommended to stay on the Raft integrated storage, how should I avoid (or what should I do to avoid) having cluster downtime due to a network error?

Thank you

2

Hi @aldy505 ,

Are you running community or enterprise? In community the recommendation is to run an odd number of nodes to avoid potential split brain scenarios. I also noticed you have nodes in two different subnets.

Enterprise supports running an even number of nodes, and supports redundancy zones.

On avoiding downtime due to network issues, performance replication is one option (enterprise feature) but this really depends on the specific network issue. Loss of an availability zone/vlan, NTP outage, DNS outage, security group update, etc all can cause issues.

You are also correct that integrated storage is the recommended storage solution.