Jira Data Center - Opinions on migration strategy

[NinjawithPyjamas]

Looking for opinions on how ppl would plan a migration from Jira standalone server to a full Jira Data Center deployment....

Im thinking there can only be 2 options:

1 - Fresh install on one node - Take XML backup from current and import to get the new database up to date. Then take the jira local directory from previous and copy over to new shared ... Then update all my conf on that one node and then move that package to my new nodes

2 - Take the jira home and local directories from the current and put them on the new. Delele the dbconf xml file and then connect that to the new database. Import the XML and move across the files for the shared directory

[Jimmy Seddon · Accepted solution]

Hi @NinjawithPyjamas,

We didn't actually end up completing the migration from Server to Data Center (We ended up going to Cloud instead). However, when we were evaluating Data Center with a test instance we followed the same path as your option 1.  I'd recommend that route since you can do some testing on an instance that doesn't affect your production server instance, then once you are confident in the setup process you can tear down that instance you tested on and perform that again when you are ready to do the live migration.

That's my two cents.

-Jimmy

[Nic Brough -Adaptavist-]

Both of those are an option, but I've got a couple of variations you might want to look into as well.

First, you talk about using the xml backup - actually, a better option is a database dump and restore.  This is faster and more reliable, and also the XML can have size problems if your data set is large (and as you are looking at DC, I suspect it is)

Obviously, that only works if you're staying with the same database - postgres -> postgres, mysql -> mysql and so on.

Second, there's a third route.  Take your existing installation and convert it to a single node Data Centre installation.  Atlassian are making this progressively easier - the current aim is to get to a point where you simply replace your Server licence with your DC licence and not have to do anything else.  The later versions of Jira are almost there, I think you still have to configure some cluster configs in a file and restart it, but it's a lot easier than it used to be.

Once it's running as a node, you can then add the other nodes you want (and eventually, drop the original if the plan was to get it off older hardware etc)

[Jimmy Seddon]

Awesome!  I wasn't aware of that!  Good to know for the future thanks @Nic Brough -Adaptavist- !

[NinjawithPyjamas]

@Nic Brough -Adaptavist-i will be going from embedded to MS SQL so that is why I was going to go with the XML option .... Option 1 is prob best since Jira install is quite simple ... Thought process is:

- Install new bits (set up DB in parallel)

- Connect to new DB

- Get UI operational and import data via XML

- Get the Jira Local Directory Data and move it across to the shared directory

Make sure nothing is missing in terms of Add Ons / Users / Cases / Attachments

[Nic Brough -Adaptavist-]

@Jimmy Seddonyep, we've used it a couple of times now.  Not a "convert production" as simply as I said there, but not far off - a flat copy of production on to a new machine, different base url temporarily, convert it to DC and then when it's looking healthy, swap it with the production machine (rollback is easier if you can just say "don't swap production out")

@NinjawithPyjamasOk, that's a good plan, I would be tempted to consider the h2 to MS-SQL migration urgently and get that done before looking at anything else.  h2 can fail catastrophically without warning (make sure you have frequent backups, of the database files in the home directory, the attachments and get it in XML so you have more recovery options)

One minor thing - you should reverse your last two steps and do less in one of them.  Once you've got your empty Jira prepared, copy only the attachments directory over from the old machine, and the XML file to import. 

When you import the XML, everything else in the home directory will be updated with stuff to support the XML, and you don't want to destroy that by copying the old data from the old machine.

[NinjawithPyjamas]

@Nic Brough -Adaptavist-great point ... So I would only copy (for example) everything in /opt/atlassian/jira_home/data/attachments/

from the OLD to the NEW instance?

[Nic Brough -Adaptavist-]

Yep, plus the xml you took for the migration.

[Nic Brough -Adaptavist-]

Argh, sorry, I forgot - you'll also want to copy the plugins directory, otherwise you'll need to re-install the plugins after migration.