Forums

Articles
Create
cancel
Showing results for 
Search instead for 
Did you mean: 

Data Centre Node Upgrade

Vickey Palzor Lepcha
Rising Star
Rising Star
Rising Stars are recognized for providing high-quality answers to other users. Rising Stars receive a certificate of achievement and are on the path to becoming Community Leaders.
July 26, 2019

When I upgrade one node of a Jira Software Data Center - with a total of 3 nodes, how does the upgrade work in terms of the database and tables.

Will it not impact views for users that are running on remaining two nodes ?

1 answer

1 accepted

1 vote
Answer accepted
Andy Heinzer
Atlassian Team
Atlassian Team members are employees working across the company in a wide variety of roles.
July 29, 2019

Hi Vickey,

I see that you are looking to understand more about upgrading Jira Data Center.   Data Center has a feature called Zero Downtime Upgrades, that is not available to Jira Server, and in turn allows you to perform the upgrade without any visible downtime to your end users.  The short answer is, yes, your end users won't notice Jira being down when performing a zero downtime upgrade. 

The longer answer, well there is a basic guide on using this feature in Managing zero downtime upgrades, but that said I hope I can explain how it works a bit further here.

Upgrading this way requires you to have at least two nodes running in your data center cluster to work properly.  But the basic idea is quite different than upgrade Jira Server.  In a single node Jira Server upgrade, Jira has to come to complete stop, there tends to be updating of application files, and upon the next startup of Jira, upgradetasks are run upon the system.  These upgradetasks usually have a fair bit of SQL updates that happen to your Jira data in the database.

Zero downtime upgrades don't work in that way.  Instead, each node will be restarted during this upgrade (one at a time) where that node will startup with a set of feature flags.  When the nodes is in the process of restarting, your load balancer is not expected to send any traffic to that node.  When that nodes comes back online, the load balancer can starting sending users to use that node again.  These restarted nodes have not yet run the upgradetasks, but indicate that this node is ready to do so.   Only once all the nodes are in this 'ready to run upgradetasks state' will you, the admin, have the option to run these.  There is more details in Zero downtime upgrades for Jira Data Center applications.

There is also a cool video of how this feature works from Summit 2016 in https://youtu.be/O4PDARm7EgQ?t=1501

It has a pretty good explanation of what happens to the nodes in this process.

I hope this helps.

Andy

Suggest an answer

Log in or Sign up to answer