Can you have multiple crowd instances for the same atlassian instances?

[P]

Meaning can we have two crowd instances used to authenticate against one jira instance? 

We just had an outage and I'm trying to think of a way to not have my current crowd server be a single point of failure. 

Is this something that can only be done with the datacenter version. 

Or can I just standup another instance. 

[Nic Brough -Adaptavist- · Accepted solution]

Yes and no.

The answer to the title question is yes - you could have several Crowd servers, each one providing a list of users (and groups, and user attributes etc) to one Jira/Confluence/Bitbucket/etc

But the short answer to the body of the question is "no".  If you created three Crowd servers, and then added "Rob" and "Nic" to all three of them, with the same login, email, location, inside-leg-measurement and so-on, then your Jira would see six different users, two per directory.  You could then do some clever stuff with alisasing, but you'd have to do it for every user, and groups wouldn't support it.

This is not a solution for resilience.

Crowd Data Center is the solution for reslience - it will look like one directory, and if one of the nodes dies, the others will carry on providing that single directory with a single list of non-duplicated users.

[P]

Thanks for explaining my options. 

[Craig Castle-Mead]

Hey Rob,

The answer from Nic doesn’t quite align with my experience (and we massively rely on Crowd). If you had two crowd servers connected to Jira, a user will be synced and authenticated from the first directory that the username exists in, there’s a connection timeout for each directory as well, so if first.last@domain.com is in the top 2 directories, directory 1 is attempted, if the timeout is reached, it’ll move on to directory 2.

As for group memberships, if a the same username is in GroupA in directory 1 and GroupB in directory 2, JIRA will place that user in GroupA and GroupB. 
 
Even though you may be able to improve your resiliency with a multi crowd instance approach, I’m with Nic, invest in doing it “properly” using Crowd DC in a multi node approach. Crowd is critical to every piece of our Atlassian stack and since we love to DC our uptime has been ~ 99.99% - the downtime we do get generally comes from the VPNs of DCs that Crowd relies on to pull its users from, and that’s not Crowds fault. 

CCM