Open john-kheir opened 6 years ago
Currently install actions in VM, Account and VDC templates have retry decorators catching BaseError. Opinion of @john-kheir is that we have to filter errors that can't be solved by retry. @FastGeert, @zaibon, could you give your input on it?
This looks like an input validation check that should not be retried, just like John says.
@katia-e
- github.com/openvcloud/0-templates/sshkey/0.0.1__59b99453:
name: '63b8de28'
dir: '/root/.ssh/'
passphrase: '59b99453'
- github.com/openvcloud/0-templates/openvcloud/0.0.1__2b32fa7f:
name: 'be-g8-3.demo.greenitglobe.com'
address: 'be-g8-3.demo.greenitglobe.com'
login: 'gig_qa_1@itsyouonline'
token: '****'
location: 'be-g8-3'
- github.com/openvcloud/0-templates/vdcuser/0.0.1__49e18fba:
openvcloud: 2b32fa7f
email: 5ce35803@test.com
name: 2200f103
groups: ['user']
provider: itsyouonline
- github.com/openvcloud/0-templates/vdcuser/0.0.1__gig_qa_1:
openvcloud: 2b32fa7f
email: dina.magdy.mohammed+123@gmail.com
name: gig_qa_1
provider: itsyouonline
- github.com/openvcloud/0-templates/account/0.0.1__ea21df8e:
openvcloud: 2b32fa7f
name: 6e31adcf
- github.com/openvcloud/0-templates/vdc/0.0.1__0815f598:
maxCPUCapacity: -673
account: ea21df8e
maxVDiskCapacity: -383
maxNetworkPeerTransfer: -247
maxMemoryCapacity: -231
name: 303f11fe
maxNumPublicIP: -122
actions:
- template: github.com/openvcloud/0-templates/vdcuser/0.0.1
actions: ['install']
- template: github.com/openvcloud/0-templates/openvcloud/0.0.1
actions: ['install']
- template: github.com/openvcloud/0-templates/vdc/0.0.1
actions: ['install']
- template: github.com/openvcloud/0-templates/account/0.0.1
actions: ['install']
@FastGeert , after discussion with @john-kheir I understand problem as follows:
Problem: Retry on install actions is triggered by batch of errors returned by OpenvCloud. Most of these errors can't be solved by retrying and need to be reported as soon as possible. For example, OpenvCloud implements number of validations that are not necessary to repeat in 0-robot templates entirely, authorization errors also should not trigger rety.
Proposal: Run retry only on errors caused by network connectivity issues, like timeouts.
After Running a blueprint to create a cloudspace with wrong prameters for instance, I'm getting these logs inside the zrobot server .. I think such kinf of errors should be filtered as i don't need to retry for a 400 bad request for example.
Zrobot server output