Our build server was transfered from Planet Federal to GeoCat hosting in 2019, with Torben and Juan taking care of the transition. Tom Ingold from Planet Federal was available to update DNS record to new server location.

In the past two years the build server performance has suffered and a number of mitigation steps performed:

• The server has suffered multiple outages due to be out of space.

• Build times for individual jobs are too long, held up primarily on disk speed.

• The release process takes hours for each stage, with recent security vulnerability roll out taking days to release.

• The number of job groups have been reduced, with more jobs being scheduled on the main node.

• Jobs have been setup with their own maven repository (to avoid corrupted jars due to shared .m2/repository use). The corruption ended up being an issue with the OSGeo Nexus repository, but each job presently has its own local maven repository.

• Each Java 11 build chain (geotools / gwc / geoserver) shares a workspace, and had previously been assigned its own execution group.

• A Docker build chain has been setup for CITE testing, but has largely been disabled due to server space issues.

• June 10th prep VM for transfer
• June 13th Purchase AX101 instance and transfer VM
• June 15th Verify builds and update DNS entry
• June 20th Restore jobs to were disabled to lack of space

1. Purchase Hezner AX101 dedicated server.

2. Install virtual box

3. Backup VM before transfer

4. Prepair VM for transfer:

   • Shutdown jenkins
   • Remove jenkins workspaces (they will be recreated)
   • Remove local m2 repositories
   • Reclaim as much space as possible.

5. Transfer VM to new server.

6. Change DNS build.geoserver.org to the new IP address

   This may be problematic as planet federal has transferred domain to OSGeo, but not released some DNS record.

7. Rollback storage restrictions

   • Restore job groups for Java 11 jobs
   • Restore CITE jobs which were shut off due to lack of space