Vault: Add active/active capability

[bkrodgers]

Per https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/vault-tool/px3M0fg3d7g/2se9LpEg2DYJ --

While I understand the complexities mentioned in the thread above, I'd like to see the ability to have multiple active nodes for HA reasons, even if scalability is not a concern. The current Active/Standby mode means there's a full outage if the active node goes down or needs to be rebooted, between the time it dies and when Vault notices and elects a new leader. I also think it would be better to be able to scale out.

[ryanking]

To deal with planned reboots of the active, a command to demote that process would be useful (can't it if it exists).

[jefferai]

@ryanking You can use vault seal to cause the active leader to give up serving client requests and force a transition. You can also just stop Vault normally as that will first seal, causing the same transition. Either of those is preferable to simply SIGKILLing the Vault process.

[seeder]

there is also step-down option on vault command to force it to drop leadership

[jefferai]

Closing for now as there are no near or long term plans for active/active.

[reverendtimm]

I think the unhappy faces on the closing comment are appropriate, to say the least.

I don't see what the issue would be with active/active?

If you're using something like DynamoDB (as in our deployment) with proper locking, why would active/active be a cause for concern. Running Vault in docker in AWS is, quite frankly, a nighmare without active/active, and you're thusly required to run it in EC2.

HA on vault seems to be quite poorly thought out. When mentioning the active/passive architecture in vault to anyone, their first response is "but why?".