Clustering and data segregation

[Bertrand Delaitre]

Hi all,

On a secure network with confidential data, we need to deploy Jira for several small projects (less than 5 users on each). Here are our constraints:

• No internet access
• Need to have up to date version
• Data for each project should be on a specific filesystem. Data should be physicaly seperated
• It should be possible to remove all the data of a project with no impact on others

Until today, we had a Jira Server for each project with the lowest tiers (10 users). Now, we have to go to Jira Data Center but it's not possible to buy one Data center for each project.

So my question is: How clustering works ?

• Do I need to buy a 500 users licence for each node?
• Is it possible to déclare a project on each node to have physical separation of data?
• Is it possible to simply remove a node from clustering to remove all the data of one project?

Any help will be appreciated.

Thanks

[Bertrand Delaitre]

Thanks for your answer. So we have to buy DC licences for each of our small projects.

Very bad news

[Nic Brough -Adaptavist-]

>So my question is: How clustering works ?

Clustering is a way to run one thing on many computers, it's main intentions are resilience (lose a node, everything carries one working as far as the users are concerned) and scaling (if you spread the usage over several nodes, you can avoid overloading a server.  Most of us use the simple model - every time a new session is started by a user, connect it to the least busy node)

So, your questions:

• Do I need to buy a 500 users licence for each node?
  • No.  You buy 500 users for a service.  You can have as many nodes as you want (be aware that Atlassian have only heavily tested on 1-4 nodes the last time I looked, but have done some testing above that, and there are people happily using more - 9 is the largest I've seen) 
• Is it possible to déclare a project on each node to have physical separation of data?
  • No.  It's one service.  The attachments are shared, and there is only one database.
• Is it possible to simply remove a node from clustering to remove all the data of one project?
  • I think the above answers that

What you're looking for here is "multi-tenancy".  Jira does not support that, other than having one separate instance per tenant.  DC can not support that in full.  You can get close on the front end, with careful set up of tenants, but some data can leak between them (field names and user ids are the two obvious ones), it's a pig to set up properly and non-existent on the back end.

In short, no.  Unless you're going to have (probably a single node) cluster for 500 people for every "tenant", DC is not going to work for you.

Sadly, Cloud would be ideal for you, apart from the "no internet" rule.

[Ravi Sagar _Sparxsys_]

Hi @Bertrand Delaitre 

Based on your constraints, yes you would need separate instances of Jira.

Ravi