- Article Type: General
- Product: Primo
Desired Outcome Goal:
To determine the proper Primo Shutdown and Start-up sequence that will minimize failover conditions, especially for MFE N+1 topologies.
Users with proper server access will occasionally unknowingly trigger a Primo failover (either FE or FE) while attempting to stop or restart Primo components.
If this happens and components are stopped before the failover is actually completed, problems can occur when Primo is restarted.
1. Stop BE
2. Stop FEs
3. Stop SEs
4. Stop Apache (if needed)
5. Stop DB (if needed)
1. Start DB
2. Start Apache
3. Start FEs
4. Start SEs
5. Start BE
BE controls FE failover.
--This means when BE is stopped first, then FEs can be stopped in any order without risk of causing FE failover.
FEs (whichever FE is MFE Master) control SE failover.
--This means if FEs are stopped next, then finally SEs can be stopped in any order without risk of causing an SE failover.
If a different sequence is used in an MFE N+1 environment,
* stopping an SE with a slice first, while FEs are running, will cause the SE indexer to failover to replace that missing slice. Slices often take several minutes to completely load before becoming searchable.
* stopping the FE that is MFE Master while the BE is running, will cause the Slave FE to failover and become the MFE Master.
During a shutdown it is very common to stop the components quickly, especially with SEs, before the failover process is completed, and this can cause problems.
This sequence can also be applied to other smaller or simpler topologies as well.
- Article last edited: 11/6/2014