Scarlet - Frequently Asked Questions

1. General Questions

1.1. Is it possible to use the Scarlet extension with all Jira versions?

No, it isn't.
Due to how Scarlet works, every Scarlet version is compatible only with a given Atlassian Jira version.

1.2. What if I want to upgrade my clustered Jira version?

For upgrading your Jira, you have to download a Scarlet version supporting the desired Jira version.
However, it is strongly suggested to keep your Jira version aligned with the latest Scarlet version.

1.3. Will Scarlet support all new versions of Jira?

Scarlet will support all major.minor versions, and the most important major.minor.minor ones.

1.4. What's the network overhead involved in the Jira cluster?

Thanks to the central server based cluster architecture, the network overhead is very low.
For more information, please take a look here.

1.5. My Jira deploy is very critical : is it possible to provide high availability for the Terracotta central server itself?

Yes, it is.
It is possible to easily change the default Scarlet configuration in order to support the deploy of the Terracotta central server in active/passive mode: while this procedure currently requires manual configurations, it will be provided off-the-shelf in future Scarlet versions.

1.6. Why there seems to be some delay between the time issues are created or modified, and the time they appear on search results?

Starting from Scarlet 1.1 beta versions, issues are asynchronously indexed: that is, they are synchronously created (modified) and stored, and then asynchronously indexed by a background process on whatever available Jira clustered node.
The noticed delay depends on the actual application load: if you notice an high delay, just add another Jira node to the cluster, in order to better distribute your application load.

2. Technical Questions

2.1. I heard that Lucene doesn't work on network filesystems : does it apply to the way Lucene indexes are shared among cluster nodes?

No, it doesn't, because Lucene indexes are stored and shared in memory, thanks to the Terracotta central server capabilities.

2.2. If Lucene indexes are stored in memory, do I have to put a lot of RAM in my server boxes?

No, you haven't. Lucene indexes are swapped on disk by the Terracotta central server, so the limit of your Lucene indexes size is actually the size of your Terracotta central server hard disk.

2.3. I want to cluster my Jira installation, but I have a lot of plugins : what I have to do?

Stateless plugins, or plugins whose state doesn't need to be shared in the cluster, will work without modifications.
If your plugins need to interact with the cluster, you will obviously have to rewrite them, probably using the Scarlet APIs described here.