Jira High Availability

[Sairam Yeturi]

We need to implement JIRA High availability on 2 windows server machines. I have not found any relevant blogs to implement the same. Is JIRA high availability feature available with the latest JIRA installer? If yes, can you please share the relevant links.

[Nic Brough -Adaptavist-]

Atlassian deprecated some of the useful documentation on doing this when Data Centre was made available.

First, JIRA does not cluster.  The best you can do for high availability is to have two machines set up with identical versions of JIRA, one being the master which everyone is using.  The second one is dormant - JIRA is installed but NOT running on it.  It is hooked up to a replicated read-only copy of the database, and the attachments directory is on a shared or replicated drive.  When something goes bang, you shut down the master JIRA, break database replication, make the secondary database the master, read/write, start JIRA, kick off a re-index, and tell your network to use that server as the master for now.  This is not high-availability to my mind, it's disaster recovery.  You can't have two active JIRA servers against one databse.

For proper HA, you need to move to JIRA Data Centre.  See https://confluence.atlassian.com/display/ENTERPRISE/JIRA+Data+Center

[Sairam Yeturi]

Hi, thanks for the response. This is tedious process, it will be in production and we cannot have disaster recovery method. We will anyways take database backup frequently.. Please let us know the best option. Should we go with single instance of JIRA or should we try Data Centre?

[Nic Brough -Adaptavist-]

If you go with a single instance, the best you can do is the DR setup described. If you need more than that, then yes, you need Data Centre

[Piotr Pawlik]

I was able to set up two JIRA servers in active-active scenario sharing same home folder and same database. I did not use Data Center - just pacemaker with bunch of other scripts etc...
The only thing Im consern is the license. Is my JIRA software license valid for such a set-up?

BRs,

Piotr

[Nic Brough -Adaptavist-]

That's hideous, you're going to have massive performance problems once it grows beyond a small system, and it will quickly come to corrupt your data.  Your index will become damaged very quickly, you're losing session information and, and, and, ...

If you persist in running a broken system like this, you are going to need TWO licences - one for each JIRA server.  But please, undo it, and run it properly.

[Piotr Pawlik]

Many thanks for pointing out mistakes in my approach!

BRs,
Piotr

[serge sterck]

No totaly agree with you

We use pacemaker with a postgres sql data in quorum with huge of trafic

The principe of pacemaker is to use un disk un sync (with drdb like a raid 1) and providing a  virtual ip

the disk in syjnc constains your database & your data.

There is one server actif so it is not a active - active configuration as said but a drs configuration and nos possibility to scale horizontaly , all must be donne with one single server.

The approach of the datacenter from atalsian relay on a load balancer like foir exemple ha_proxy and a share nfs but could also be a DBDR raid 1 replication like pacemaker does. This aproache is better because no single of failure of your nfs server.

The only advantage is active active and to possibility to spread your traif on multi server horizontanly..

For the license there are not stand by license like IBM products , the jira take a footprint of your server. So when you are goind to disaster regenerate your

key for your disaster server