changing data residency

[Darren Cornelius]

I am looking to move all products to EU but the general guidance is poor in terms of time offline along with search capabilities. 

This is all dependent on size - but cannot see any kind of guides so I can set realistic expectations of how long the system would be offline.  Can anyone shed any light on this ?

[Evgenii]

Hi, @Darren Cornelius 

Analyzed information, from Atlassian KB about Downtime Expectations:

Initial Migration Downtime:

• Downtime may last up to 24 hours, depending on data size Atlassian Support

• Best practice is Atlassian Support to plan for your app to be offline for up to 24 hours

Move Window Options:

1. Maintenance window - Available if estimated downtime is under 2 hours

2. Within 48 hours - Available if estimated downtime is less than 20 hours

3. Scheduled custom time - You choose the date/time, but must request at least 3 days in advance

The Critical Search Issue

Search functionality may be temporarily disabled during re-indexing, which can take up to 3 days depending on your data size Atlassian Support.

This means:

• Your products come back online after the migration (2-24 hours)

• BUT search may not work properly for up to 3 additional days during reindexing

• Users can access the system, but search/JQL queries may be incomplete or unavailable

[Darren Cornelius]

yeah it was this broad - migration (2-24 hours) - that scares me ...

[Evgenii]

I understand, why Atlassian are writing "up to 24 hours", they don't want complains. 
In my experience - 10 Gb migrated during a an 2 hours (even faster). It's not a very large amount of data. It was made before weekends, so users didn't noticed anything.

[Marc -Devoteam-]

Hi @Darren Cornelius 

You can check your amount of storage in Confluence.

If this is high I can tell you the move can take a long time.

To be sure on expected time, best is to raise a request at Atlassian Support

They are able to give you an indication on who long the move would approximately take

[Darren Cornelius]

tried logging a ticket but due to the way these are now categorised - it was closed as didn't match ! couldn't find any suitable area to log it against.

roughly 8GB JIRA and same in confluence so doesn seem large amounts ..

[Marc -Devoteam-]

Hi @Darren Cornelius 

Based on this amount of data, as mentioned by @Evgenii this migration should not take too much time.

So scheduling the move in the weekend, this would give Atlassian and you enough time to execute and check the instance after the move.

[Kris Klima _K15t_]

Hi @Darren Cornelius 

From experience, I haven't noticed anything happend.

My private site is not big and not critical. But still, I think it happened overnight because I never experience any glitches (except one which was specific to a single app and has now been rectified).

Same on on my previous job. Again, it happened overnight, and we would not notice. Our public docs were created by Scroll Viewport / Sites which was independent of Confluence... so knowing there'd be no disruption for out end users kinda caused me to pay less attention. 

What I'm seeing now on my private instance is that move was delayed. I guess there's a lot of requests these days... 

Server downtimes for maintenance were much more severe.

[Darren Cornelius]

our site is critical as using JSM operations and providing 24x7 support. Not large amounts of data only 8GB per site 

[Tomislav Tobijas]

Hey @Darren Cornelius ,

I did a couple of Data Residency moves and they all switched relatively fast. Actually, the fastest times I've seen are if you would schedule the move within the next 48 hours.

What does estimated downtime say when you start planning the move?

In my cases, everything that said up to 2 hours moved in under a couple of minutes. And we also mainly moved data to the EU and once to the US, and some were bigger than what you described you have.

⚠️ But that was only my experience > meaning, takes much less than estimated, with even general applications being up and potentially only some functions not working for that time window.

I'll have to agree with you that docs don't provide much support here and these estimations seem more like 'worst case scenarios' than actual required time.

Cheers,
Tobi