Deadlock when job in serial_group fails #163
Description
We have three jobs that are in the same serial group. The second job triggers once the first passes, and the third triggers when the second passes. On our first run, the first job failed. The other two jobs showed that they were pending. When we tried to start the first job again, it got stuck in the pending state. We tried aborting it and creating a new one, but that too got stuck in pending. We were able to get it to start by aborting the other jobs.
jobs:
- name: superman-cf-deploy
serial: true
serial_groups:
- superman-deployment-group
plan:
- aggregate:
- get: deployments-routing
- get: cf-release-rc
trigger: true
- get: bosh-stemcell
...
- name: superman-diego-deploy
serial: true
serial_groups:
- superman-deployment-group
plan:
- aggregate:
- get: bosh-stemcell
- get: deployments-routing
- get: cf-release-rc
passed:
- superman-cf-deploy
trigger: true
params:
submodules:
- src/loggregator
- get: diego-release
trigger: true
...
- name: superman-routing-deploy
serial: true
serial_groups:
- superman-deployment-group
plan:
- aggregate:
- get: deployments-routing
- get: cf-routing-release-rc
trigger: true
- get: diego-release
trigger: true
params:
submodules: none
passed:
- superman-diego-deploy
...