Do small batches of each type of server before doing larger batches
Similar to the rationale on !4 (closed), we change our batches so that we can test servers of different type/age so we can detect possible failures related to particularities of each node type in a more limited blast radius. In this proposal we keep each of the initial small batches homogeneous so if an error arises we can focus on a single type of node at a time. After executing small batches of each node type and verifying they work correctly, we continue with larger batches for the rest of the nodes of each type.
| Batch | Name | No. servers |
|---|---|---|
| 1 | part single zone | 4 |
| 2 | hdd | 4 |
| 3 | marquee | 4 |
| 4 | pre 2019 | 4 |
| 5 | post 2019 | 4 |
| 6 | hdd and part multizone | 21 |
| 7 | multizone | 24 |
| 8 | rest of single zone | 16 |
| 9 | praefect | 3 |
Edited by Rachel Nienaber