search

xchem / xchem_it

Issues for XChem IT work
0 stars 0 forks source link

Migrate storage to Longhorn #23

Open tdudgeon opened 3 years ago

tdudgeon commented 3 years ago

Longhorn is deployed to the Dev cluster (see #12). Many of the volumes that are currently in use are probably best switched to Longhorn as storage will be local to the nodes and as backups can easily be managed. Use of Longhorn for read-only-many and read-write-many volumes has not yet been tested, but assuming these work as documented then Longhorn should probably replace all use of NFS volumes. The postgres database volumes are other good candidates for switching, but will require a backup and restore of the db data.

Cinder might still be needed, especially where very large volumes are needed (e.g. graph dbs), though large Cinder volumes could be added to Longhorn.

When considering each volume the following should be considered:

phraenquex commented 3 years ago

Impact of not doing: currently, if postgresql crashes, there will be more (lots of) downtime for getting it back up. With Longhorn, it would do fail-over immediately.

tdudgeon commented 3 years ago

Immediate actions:

  1. Migrate the media volume in Duncan's stack to Longhorn to establish that Longhorn can be used for an entire stack.

Medium term actions:

  1. Migrate the postgress volume in the staging stack from NFS to Longhorn (NFS is not recommended for PostgreSQL, but it was all that was available at the time).
  2. ditto for the production stack
  3. migrate the other developer stacks

Longer term actions:

  1. consider all remaining NFS and Cinder volumes to see if they should be migrated to Longhorn (Cinder will remain best option for some).