re: Jira Data Center and clustering between different geographical locations.

[Bryan Karsh]

Hi guys,

I noticed in the FAQs for JIRA data center this question:

How can I deploy JIRA Data Center in different geo-locations, such as Americas, Europe, and APAC?
JIRA Data Center is designed to support hot failover scenarios in a co-located data center. Geo-clustering is not supported in JIRA Data Center today.

We have silos in different US locations. If we clustered Jira, we would want to spread the nodes between these locations, in the event that one of our silos went down. Can JIRA data center support this? Or can it only support HA within a single "Availability Zone" ?

[Joe Clark · Accepted solution]

Hi @Bryan Karsh,

The individual nodes in a JIRA Data Center deployment are designed to communicate with each other in a high-bandwidth, low-latency environment.  The best way to do this is to co-locate all the nodes in the same place.

Locating the nodes in geographically diverse locations puts big restrictions on the bandwidth and latency that can be feasibly achieved.  If the nodes are too "far" apart (in a virtual sense), they will be constantly panicking and struggling to stay synchronized with each other - the nodes will appear to go offline and come back online erratically.  The end result is a cluster of nodes that end up with worse performance than a single node operating on its own.  As such, it's not recommended to deploy JIRA Data Center in this kind of configuration.

If you want to protect your JIRA deployment against geographical outages, then you could look at a cold failover to a different availability zone. 

 This may change in the future as JIRA Data Center matures and becomes able to handle a wider spectrum of availability requirements.

[Bryan Karsh]

Thanks -- that makes complete sense. Appreciate the quick reply.

[Bryan Karsh]

Hi -- a follow-up question. Would multiple AWS availability zones within the same region work, latency-wise? Or what about 2 data centers in the US about 1000 miles apart?

[Bryan Karsh]

Another question -- what about a shared cross-site NFS mount? Some of team thinks they could set one up between 2 of our datacenters. Is that feasible, and if not, can you explain? I'm assuming similar latency issues mentioned above?