Logic to generate cluster wide uniq sequences
Why?
To finalize the logic to generate unique database sequences across cells. We have discussed the below solutions in the past, have to see if we can add have more possible solutions and finalize the best one.
- Having cell specific info (eg: cell-id) in the MSB bits of the sequence.
- Having timestamp and cell specific info.
Also we have to ensure that the primary cell's sequences should not interfere with the secondary ones.
Solutions in consideration:
- Solution 1: Topology service issues range of sequences for each cell
- Solution 2: Using timestamp to generate uniq IDs
- Solution 1.5: Unique sequence per cell with topology service, use simple integer arithmetic
Finalized solution:
A revised version of the solution 1 has been finalized.
Deliverable
-
Finalize a solution -
Update the cells blueprint with new approach, and consolidate the changes in https://docs.gitlab.com/ee/architecture/blueprints/cells/iterations/cells-1.0.html#primary-cell and https://docs.gitlab.com/ee/architecture/blueprints/cells/topology_service.html#sequence-service
Edited by Prabakaran Murugesan