JIRA / Confluence Data Center (AWS) Best Practices

[Albert Manya]

We are starting to work in setting up a JIRA and Confluence configuration as much high available as possible. We will use AWS to do so and will use the Data Center versions.

The idea is to start with 500 users but scale up to around 2000 or more.

We are wondering what good practices are people using to do so, for example:

• For the shared storage, we wanted to be distributed and ha so NFS doesn't applies, what are you using? GlusterFS? Ceph? We are also thinking on s3 fuse but are afraid of the performance.
• For the database we are going to use RDS, somebody have tried?
• For the load-balancer we are going to use ELB
• Somebody uses some sort of auto-scaling or its done by hand.

Thanks!

[Corentin Méhat]

Hello,

since this question Atlassian published the following pages about Datacenter deployment on AWS :

https://confluence.atlassian.com/adminjiraserver072/running-jira-software-data-center-in-aws-857064325.html

and

https://confluence.atlassian.com/adminjiraserver072/recommendations-for-running-jira-software-data-center-in-aws-857064327.html

I personnaly succesfully set up a JIRA DataCenter on AWS a bit before they published their guide with the following components :

DB : RDS Mysql
Servers : EC2 instances
Load balancers : ELB
Shared FS : nfs
hope it helps :)

[jameers]

Hi 

Corentin MEHAT

Thanks, now links are not working. can you please explain it. can we install nfs in the same instance or we need separate nfs server?

 DB : RDS PostgreSQL
Servers : EC2 instances
Load balancers : ELB

NFS:?

[Corentin Méhat]

Now that it is globally available I would recommend using EFS for the shared folder rather than supporting your own NFS installation.

[Ahmad Awadallah]

I just came back from re:Invent.  I learned about auto-scaling and also would like info in this area for JIRA and Confluence.  I guess it also depends on what type of auto-scaling you want to use... spinning up more instances or containers.  This is what I am trying to find info about.  Either way does Atlassian support multiple instances hitting the same DB?  And what about the DB... this looks like it may still be a bottleneck in performance since Atlassian does not support NoSQL DBs.  

If you find info please let us know.  If I find something useful I will come back here and post it.

Thanks.

[Tran Dien Hien]

Hi All,

I have been deploying Confluence Data center on AWS bas below:

-EC2 Node auto scaling

- ABL : Application load balancer

- EFS: Share home

- DB : Selfhost

But I'm facing a problem with the cluster, when when the auto scaling increase more node the cluster don't work because  because new node have the same cluster id with the current node. How can I solve this problem? Have any solution to change the cluster node automatically.

I much appreciate for your help 

[Chris Moreira]

Hi Albert, did you end up figuring this out? please share?

[Kevin_Terminella]

We've been testing out GlusterFS and it seems to behave well both from a functional and performance perspective.

[Chris Moreira]

Hey Kevin, so have got JIRA datacenter working on the AWS? how about the Multicast problem, was that an issue?

[Kevin_Terminella]

Wasn't a problem as you can choose to not use multicast for JDC.

[Chris Moreira]

thx that is great news... so how many nodes u running? about the GlusterFS performance, would you be able to share how many actives user you have roughly?

[Kevin_Terminella]

We are testing against 2000-10000 TPM, with a lot of the transactions centered around viewing and updating issues. EFS was just released today, so that may be a better option.

[Kevin_Terminella]

set ehcache.peer.discovery = default. The nodes get registered in the database and that's how other nodes know their location. The multicast option (ehcache.peer.discovery = default) is no longer recommended by Atlassian.

[Dave Theodore [Coyote Creek Consulting]]

GlusterFS is not supported by Atlassian, so you may not want to go this route. We've implemented quite a few Data Center environments of all applications in AWS and on self-hosted virtual environments.  We've found that EFS has good performance and is suitable for Data Center.  For self-hosted, we've seen NFS work great and we've seen it be nearly unusable.  You really need to measure the storage latency to make sure it's fairly low.  We don't recommend using unsupported technologies, since they will not have been tested by Atlassian and they will have a challenge in helping you if your support issue may be related to storage or storage performance.

[Chris Moreira]

Hey guys, did anyone find the solution for this?  I'd appreciate if you could share that 

[Everett Daviage]

We are looking for similar answers. Has anyone contacted Atlassian directly with this question?

[Kevin_Terminella]

Did you end up figuring out a solution? I'm trying to design JDC on AWS as well and keep getting hung up on the shared storage piece.