488c99dab3
The following potential problems were observed with FrozenJob secrets: 1) They may be repetitive: since the FrozenJob contains lists of playbooks and each playbook record has a copy of all the secrets which should be used for that playbook, if a job has multiple playbooks the secrets will be repeated for each job. Consider a base job with three playbooks: the base job's secrets will be included three times. 2) They may be large: secrets in ZK are stored encrypted and suffer the same size explosion that they do when encrypted into zuul.yaml files. 3) Take #1 and #2 together and we have the possibility of having FrozenJob objects that are larger than 1MB which is a problem for ZK. Address all three issues by offloading the secrets to a new ZK node if they are large (using the existing JobData framework) and de-duplicate them and refer to them by index. There is no backwards compatability handling here, so the ZK state needs to be deleted. Change-Id: I32133e8dd0e933528381f1187d270142046ff08f
10 lines
220 B
YAML
10 lines
220 B
YAML
---
|
|
upgrade:
|
|
- |
|
|
Due to a change in the ZooKeeper data format, the following
|
|
upgrade procedure is required:
|
|
|
|
* Stop all Zuul components
|
|
* Run ``zuul delete-state``
|
|
* Start all Zuul components
|