sfctl sa-cluster config-upgrade returns invalid argument

[Christina-Kang]

See MicrosoftDocs/azure-docs#18835

When reporting a new issue include the following information:

• sfctl version, can check this using pip show sfctl
• Service Fabric cluster runtime version
• Service Fabric hosting location (on-premise or Azure usually)
• Debug log if an error is present (can be captured using --debug flag)

Be sure to check if an issue already exists with your symptoms. If so, leave a :thumbsup: reaction to the original issue, rather than creating a duplicate.

[krvladislav]

Sfctl version = 6.0.0
OS = Windows 10 version 1803 (OS Build 17134.228)
Service Fabric cluster runtime version = 6.3.187.9494
Service Fabric hosting location = on-premise Windows Server version 1803 (OS Build 17134)
Version of Python = 3.7.0
Debug log: https://hastebin.com/geheteheye.makefile

I sniffered request from sfctl:

{ "ClusterConfig": "<CONFIG>", "HealthCheckRetryTimeout": "P0D", "HealthCheckWaitDurationInSeconds": "P0D", "HealthCheckStableDurationInSeconds": "P0D", "UpgradeDomainTimeoutInSeconds": "P0D", "UpgradeTimeoutInSeconds": "P0D", "MaxPercentUnhealthyApplications": 0, "MaxPercentUnhealthyNodes": 0, "MaxPercentDeltaUnhealthyNodes": 0, "MaxPercentUpgradeDomainDeltaUnhealthyNodes": 0 }

Now I make raw REST request without timeout fields and E_INVALIDARG error disappears:

{ "ClusterConfig": "<CONFIG>", "MaxPercentUnhealthyApplications": 0, "MaxPercentUnhealthyNodes": 0, "MaxPercentDeltaUnhealthyNodes": 0, "MaxPercentUpgradeDomainDeltaUnhealthyNodes": 0 }

[krvladislav]

"P0D" is valid representation of duration by ISO 8601, maybe problem is in SF REST API.

As a workaround I pass own 1 second timeout, but I don't understand real difference under the hood with default value.

$timeout = "PT1S"
sfctl sa-cluster config-upgrade --cluster-config $clusterConfiguration --health-check-retry $timeout --health-check-wait $timeout  --health-check-stable $timeout  --upgrade-domain-timeout  $timeout --upgrade-timeout $timeout