Some Confluence and Jira cloud users may experience brief interruption due to scheduled maintenance

[#1]

Database upgrade scheduled for Confluence and Jira cloud sites

The current version of PostgreSQL (version 13.9) used by Confluence cloud and all Jira cloud products, will reach end of support from AWS in March 2024. In order to continue to provide reliable service to customers, we will be upgrading the database to PostgreSQL version 13.12.

This infrastructure upgrade will begin rolling out to customer sites from 29 February 2024 at 12AM (UTC), and is scheduled to be complete for all users by 15 March 2024 at 1PM (UTC). We understand the importance of uninterrupted service, and we expect only minimal downtime lasting up to 2 minutes for users. We expect the majority of users will not notice the interruption.

Will the upgrade take place during my defined maintenance window?

Unfortunately we cannot upgrade sites or products individually on a per customer basis. The PostgreSQL upgrade can only be applied at the database level and as such applies to an entire cloud hosting region at once. For example, all customer sites hosted in the eu-central-1 region must be upgraded at the same time.

As a result, there is no guarantee that the upgrade will take place during each customer’s defined maintenance window. However, in order to minimize disruption to the vast majority of users, the upgrade will occur between 1am and 3am within the local timezone of each region being upgraded. This window reflects the period where we see the least user activity and is least likely to cause interruptions.

What users may experience during an upgrade

Rest assured, our team has extensively tested this upgrade on our own production Atlassian site, and interruptions have been negligible. At most a user may experience up to 2 minutes of downtime, similar to a brief internet outage.

Any users in the middle of entering data during the upgrade will temporarily be unable to edit or publish, but autosave features will function as normal to retain drafted content.

In rare cases, users may temporarily see a '500 Error' while performing an action.

Impact on Migrations (added 29 February 2024)

The majority of our cloud migration assistant tooling are built with re-try mechanisms that will attempt to migrate your data at least 5 times prior to failing. Each re-try runs for longer than the 2 minutes of downtime caused by this upgrade work, and so the migration will be able to continue without needing to restart. This may add additional time to a migration by a few minutes, but overall should have minimal impact on customer migrations.

This retry mechanism is included in the following migration methods:

• All Confluence migrations via CCMA (Confluence Cloud Migration Assistant)

• All Jira migrations using the recommended "Choose what you migrate" method via JCMA (Jira Cloud Migration Assistant)

This retry mechanism is not included for Jira migrations using the "Migrate all at once" (MAO) method via JCMA. In the event that the downtime occurs during a Jira MAO migration, the import may need to be manually re-run.

If you are concerned that your migration may occur at the same time as an upgrade, please contact us through your existing migration support channel if you have one open, or raise a new ticket with migrations support via https://support.atlassian.com/contact/

Our commitment to you

We are committed to ensuring the highest level of service, performance and reliability for your Confluence and Jira platforms. While we cannot change the timing of when the upgrade will occur for you and your products, if you have any further questions or concerns regarding this upgrade, leave a comment here or reach out via our support channels.

[Andy Gladstone]

@Lauren Button thank you for this detailed bulletin. Although the downtime will be minimal, many companies would choose not to communicate at all and just 'deal with it' when users report outages. Good on Atlassian for keeping true to their core values and providing transparency to its users and subscribers.

[Anthony Weicker]

I agree with @Andy Gladstone , transparent over communication is fantastic!

Is there a way to be notified closer to our particular maintenance window, and then also once the upgrade is complete?

[Béla Hegyesi]

Hi Lauren, 
Thanks for the detailed description.
May we ask, why are you just performing a minor version upgrade instead of going in a more future-proof direction?
According to AWS's documentation, this version will be supported only for a few additional months.

[Christian Schneider]

Dear Lauren,

thank you for the notice in advance. I received an email, saying that there will be a downtime of 2min within the next 336 hours. Thank you for the transparency, but what do you expect me to do with this bit of information? This is not even worth communicating to my team, as there is nothing they can prepare for.

Just go on with your upgrade and make sure, it does not blow up everything. :-)

Chris

[Lauren Button]

Hey everyone! Thanks for the comments. I'll try and address some of your questions:

Why are we notifying customers about such a small impact?

As many customers have tailored their maintenance window outside of our default window, there's a chance that this upgrade could occur outside of each customer's individual maintenance window. So we wanted to be transparent about this scheduled downtime although yes, it is small!

Is there a way to be notified closer to our particular upgrade / be notified once our upgrade is complete?

Given this impacts all Confluence and all Jira product customers, it is tricky to coordinate that level of personalised communication given the exact schedule of upgrades may change as we progress through this rollout. We do hope in future to build more automation into the process in order to provide more tailored, specific communication. I'll share this feedback with the team.

Why are we performing a minor upgrade rather than a major upgrade?

Let me check in with the team - I know there was a decision paper on which version we would upgrade to and both minor and major versions were considered before we landed on v13.12. I'll report back once I have more info!

[Sascha Metz]

Hello @Lauren Button ,

do you ensure the backup strategy or do we have to somehow prepare backups and export vital Confluence articles and the active sprint board? As learned in the past preparation is normally not needed but useful in exceptional situations like https://access.status.atlassian.com/incidents/6lk9kcd9h3hd

Best regards & a good start into the weekend

[Lauren Button]

Hey everyone, sorry for the delay in getting back about the decision to upgrade to 13.12 vs a more recent version.

The main driver was wanting a direct minor upgrade path from 13.9 to another 13.X version.

We test our products with new versions well in advance, and when we first began testing for this upgrade, 13.12 was the highest available minor upgrade target.

By the time 13.13 was announced on Dec 21 2023, we had completed most prep work for 13.12, and we were also only just over 2 months away from 13.9 end of support (March 2024).

We are aiming to perform a major upgrade to a newer version over the next year to give us a longer runway in future.

Hope that makes sense!

cc @Béla Hegyesi