I haven’t found anything in the documentation that talks about the startup or shutdown order if I have all of the appliances deployed and connected with each other. Is there a prefered order, or one that makes the most sense for the system?
You’d probably want to just apply general good practices here and 1) gracefully evacuate users and 2) take steps to stop writing to things that store data before you shut them down.
- Get users off of ECA (if used)
- Shut down ECA (which will clear out any users who didn’t listen in step 1 )
- Get users off of EDA (if users are logged in directly)
- Shut down EDA so it stops writing to any EXA
- On the assumption EDA and ETA are getting a matched packet feed and you want their data to be in sync, shut down ETA “soon” after EDA
- If extended storage units (ESUs aka JBODs) are deployed behind ETA, power them off now that the ETA is no longer writing to them
- Shutdown the EXA cluster
This assumes a fairly straight hierarchy where EXAs and ETAs aren’t “cross-shared” with other EDAs and ECAs that you don’t plan to shut down. But the same principles should guide you through those scenarios.