How do we make Crowd more resilient?

Question

Our team use Crowd to control access to all our Atlassian tools. At the moment it is pretty flaky and frequently unavailable, which breaks JIRA, Confluence, Bitbucket, all those useful things.  We have offices in several timezones so sometimes that can cause long downtimes until the support team handling Crowd can be called in.

The Crowd issue is being dealt with via Atlassian Support, but I was wondering if there is a better way to keep the tools available even when Crowd is uncontactable for any reason, whether due to Crowd, netowrking or whatever.

Is there any sort of caching or local server that could be set up to allow a user to keep going provided they have authenticated against the central Crowd server recently?

Bruno Vincent · Accepted Answer

Hi @Paul Scott

Regarding Crowd's availability you might want to take a look at Crowd Data Center, which is a clustered version of Crowd: https://www.atlassian.com/software/crowd/enterprise/data-center

Marcin Kempa · Accepted Answer

Hi @Paul Scott

As @Bruno Vincent suggested, Crowd DC was build for high availability so if you look into removing single point of failure from your infrastructure it is the way to go.

However, as you described, you seem to be running into some networking issues and I believe you are looking into some solution that would allow you to build something similar to what Bitbucket Server does with repo mirroring.

Unfortunately at the moment we do not support this model of caching with local mirror instances of Crowd.

I am not sure what kind of issues you are running into exactly, are those problems between Atlassian tools and Crowd itself or Crowd and remote directories which are unavailable because of some network issues? If this is the case you could also try directory failover option in Crowd as described here.

Also, maybe taking advantage of 'remember me' cookies would alleviate your problems?

Best Regards,

Marcin Kempa