As part of our ongoing efforts to improve the Jupyter and Analytics Client user experience , we will enable shared home directories on the stats servers.
Creating this ticket to track all sub-tasks for Phase 1.
The following is a list of anticipated steps required.
- Create a ceph file system named home with data stored on HDD and metadata stored on SSD
- Create a suitable cephx client user, with rights appropriate to access this file system
- Determine what quota settings will be appropriate for this file system
- Distribute the cephx key to a test server
- Mount the home volume to /home on ml-lab1002
- Add puppet management of the /home mount point and cephx key
- Verify that puppet can create and populate the user's home directories correctly
- Ensure that we have adequate alerting and performance analysis available via dashboards
- Perform some performance and resilience testing of the home directories
- Mount /home on a single stat server - informing users that their existing homes are still available in /srv/home
- Assess user experience and solicit feedback on roll-forward roll-back options
- Assuming feedback is positive, continue to mount /home on the remaining stat servers and ml-lab1001