During backend initialization, especially during a migration, there is a
chance that an existing state could be overwritten.
Attempt to get a locks when writing the new state. It would be nice to
always have a lock when reading the states, but the recursive structure
of the Meta.Backend config functions makes that quite complex.
* Enable remote s3 state support for assume role
- provide role_arn in backend config to enable assume role
Fixes#8739
* Check for errors after obtaining credentials
Allows redshift security group tests to better handle being ran in parallel.
```
$ make testacc TEST=./builtin/providers/aws TESTARGS='-run=TestAccAWSRedshiftSecurityGroup_'
==> Checking that code complies with gofmt requirements...
go generate $(go list ./... | grep -v /terraform/vendor/)
2017/02/09 10:40:25 Generated command/internal_plugin_list.go
TF_ACC=1 go test ./builtin/providers/aws -v -run=TestAccAWSRedshiftSecurityGroup_ -timeout 120m
=== RUN TestAccAWSRedshiftSecurityGroup_importBasic
--- PASS: TestAccAWSRedshiftSecurityGroup_importBasic (12.98s)
=== RUN TestAccAWSRedshiftSecurityGroup_ingressCidr
--- PASS: TestAccAWSRedshiftSecurityGroup_ingressCidr (11.02s)
=== RUN TestAccAWSRedshiftSecurityGroup_updateIngressCidr
--- PASS: TestAccAWSRedshiftSecurityGroup_updateIngressCidr (32.81s)
=== RUN TestAccAWSRedshiftSecurityGroup_ingressSecurityGroup
--- PASS: TestAccAWSRedshiftSecurityGroup_ingressSecurityGroup (14.82s)
=== RUN TestAccAWSRedshiftSecurityGroup_updateIngressSecurityGroup
--- PASS: TestAccAWSRedshiftSecurityGroup_updateIngressSecurityGroup (37.43s)
PASS
ok github.com/hashicorp/terraform/builtin/providers/aws 109.090s
```
Allows the redshift parameter group acceptance tests handle being ran in parallel better
```
$ make testacc TEST=./builtin/providers/aws TESTARGS='-run=TestAccAWSRedshiftParameterGroup_'
==> Checking that code complies with gofmt requirements...
go generate $(go list ./... | grep -v /terraform/vendor/)
2017/02/09 10:16:19 Generated command/internal_plugin_list.go
TF_ACC=1 go test ./builtin/providers/aws -v -run=TestAccAWSRedshiftParameterGroup_ -timeout 120m
=== RUN TestAccAWSRedshiftParameterGroup_importBasic
--- PASS: TestAccAWSRedshiftParameterGroup_importBasic (15.17s)
=== RUN TestAccAWSRedshiftParameterGroup_withParameters
--- PASS: TestAccAWSRedshiftParameterGroup_withParameters (13.16s)
=== RUN TestAccAWSRedshiftParameterGroup_withoutParameters
--- PASS: TestAccAWSRedshiftParameterGroup_withoutParameters (12.58s)
PASS
ok github.com/hashicorp/terraform/builtin/providers/aws 40.940s
```
Updates the aws_elb acceptance tests to better handle parallel test runs
```
$ make testacc TEST=./builtin/providers/aws TESTARGS='-run=TestAccAWSLoadBalancerPolicy_'
==> Checking that code complies with gofmt requirements...
go generate $(go list ./... | grep -v /terraform/vendor/)
2017/02/09 10:04:58 Generated command/internal_plugin_list.go
TF_ACC=1 go test ./builtin/providers/aws -v -run=TestAccAWSLoadBalancerPolicy_ -timeout 120m
=== RUN TestAccAWSLoadBalancerPolicy_basic
--- PASS: TestAccAWSLoadBalancerPolicy_basic (24.50s)
=== RUN TestAccAWSLoadBalancerPolicy_updateWhileAssigned
--- PASS: TestAccAWSLoadBalancerPolicy_updateWhileAssigned (42.34s)
PASS
ok github.com/hashicorp/terraform/builtin/providers/aws 66.866s
```
When ran in parallel the tests `TestAccAwsEcsTaskDefinition_withNetwork` and `TestAccAwsEcsTaskDefinition_withTask` will overlap with each other due to the shared naming of the `iam_role` resource.
This fixes these tests to allow running in parallel on TeamCity.
```
$ make testacc TEST=./builtin/providers/aws TESTARGS='-run=TestAccAwsEcsTaskDefinition_withTask'
==> Checking that code complies with gofmt requirements...
go generate $(go list ./... | grep -v /terraform/vendor/)
2017/02/09 09:20:03 Generated command/internal_plugin_list.go
TF_ACC=1 go test ./builtin/providers/aws -v -run=TestAccAwsEcsTaskDefinition_withTask -timeout 120m
PASS
ok github.com/hashicorp/terraform/builtin/providers/aws 0.022s
```
```
$ make testacc TEST=./builtin/providers/aws TESTARGS='-run=TestAccAwsEcsTaskDefinition_withNetwork'
==> Checking that code complies with gofmt requirements...
go generate $(go list ./... | grep -v /terraform/vendor/)
2017/02/09 09:21:10 Generated command/internal_plugin_list.go
TF_ACC=1 go test ./builtin/providers/aws -v -run=TestAccAwsEcsTaskDefinition_withNetwork -timeout 120m
PASS
ok github.com/hashicorp/terraform/builtin/providers/aws 0.026s
```
Fixes the `TestAccAwsAPIGatewayMethod_customauthorizer` acceptance test which would previously fail if the iam_role resources would leak
```
$ make testacc TEST=./builtin/providers/aws TESTARGS='-run=TestAccAwsAPIGatewayMethod_customauthorizer'
==> Checking that code complies with gofmt requirements...
go generate $(go list ./... | grep -v /terraform/vendor/)
2017/02/09 09:10:07 Generated command/internal_plugin_list.go
TF_ACC=1 go test ./builtin/providers/aws -v -run=TestAccAwsAPIGatewayMethod_customauthorizer -timeout 120m
PASS
ok github.com/hashicorp/terraform/builtin/providers/aws 0.022s
```
An AWS Security Group Rule requires at least one of `cidr_blocks`, `self`, or `source_security_group_id` in order to be successfully created.
If the `aws_security_group_rule` doesn't contain one of these attributes, the AWS API will still return a `200` response, and not report any error in the response.
Example response from the API on a malformed submission:
```
2017/02/08 16:04:33 [DEBUG] plugin: terraform: -----------------------------------------------------
2017/02/08 16:04:33 [DEBUG] plugin: terraform: aws-provider (internal) 2017/02/08 16:04:33 [DEBUG] [aws-sdk-go] DEBUG: Response ec2/AuthorizeSecurityGroupIngress Details:
2017/02/08 16:04:33 [DEBUG] plugin: terraform: ---[ RESPONSE ]--------------------------------------
2017/02/08 16:04:33 [DEBUG] plugin: terraform: HTTP/1.1 200 OK
2017/02/08 16:04:33 [DEBUG] plugin: terraform: Connection: close
2017/02/08 16:04:33 [DEBUG] plugin: terraform: Transfer-Encoding: chunked
2017/02/08 16:04:33 [DEBUG] plugin: terraform: Content-Type: text/xml;charset=UTF-8
2017/02/08 16:04:33 [DEBUG] plugin: terraform: Date: Wed, 08 Feb 2017 21:04:33 GMT
2017/02/08 16:04:33 [DEBUG] plugin: terraform: Server: AmazonEC2
2017/02/08 16:04:33 [DEBUG] plugin: terraform: Vary: Accept-Encoding
2017/02/08 16:04:33 [DEBUG] plugin: terraform:
2017/02/08 16:04:33 [DEBUG] plugin: terraform: 102
2017/02/08 16:04:33 [DEBUG] plugin: terraform: <?xml version="1.0" encoding="UTF-8"?>
2017/02/08 16:04:33 [DEBUG] plugin: terraform: <AuthorizeSecurityGroupIngressResponse xmlns="http://ec2.amazonaws.com/doc/2016-11-15/">
2017/02/08 16:04:33 [DEBUG] plugin: terraform: <requestId>ac08c33f-8043-46d4-b637-4c4b2fc7a094</requestId>
2017/02/08 16:04:33 [DEBUG] plugin: terraform: <return>true</return>
2017/02/08 16:04:33 [DEBUG] plugin: terraform: </AuthorizeSecurityGroupIngressResponse>
2017/02/08 16:04:33 [DEBUG] plugin: terraform: 0
2017/02/08 16:04:33 [DEBUG] plugin: terraform:
2017/02/08 16:04:33 [DEBUG] plugin: terraform:
2017/02/08 16:04:33 [DEBUG] plugin: terraform: -----------------------------------------------------
```
This previously caused Terraform to wait until the security_group_rule propagated, which never happened due to the silent failure.
The changeset ensures that one of the required attributes are set prior to creating the aws_security_group_rule.
Also catches the error returned from the retry function. Previously the error was ignored, and only logged at the `DEBUG` level.
Previously, an `aws_rds_cluster` that contains active instance groups would timeout on a destroy, if the destroy was able to only target the rds_cluster and not include the instance groups.
This would result in a `400` response from AWS, and Terraform would sit in a wait-loop until a 15-minute timeout while waiting for the cluster to be destroyed.
This catches the error returned from the `DeleteDBCluster` function call such that the proper error case can be returned to the user.
`400` from the AWS API:
```
2017/02/08 13:40:47 [DEBUG] plugin: terraform: ---[ RESPONSE ]--------------------------------------
2017/02/08 13:40:47 [DEBUG] plugin: terraform: HTTP/1.1 400 Bad Request
2017/02/08 13:40:47 [DEBUG] plugin: terraform: Connection: close
2017/02/08 13:40:47 [DEBUG] plugin: terraform: Content-Length: 337
2017/02/08 13:40:47 [DEBUG] plugin: terraform: Content-Type: text/xml
2017/02/08 13:40:47 [DEBUG] plugin: terraform: Date: Wed, 08 Feb 2017 18:40:46 GMT
2017/02/08 13:40:47 [DEBUG] plugin: terraform: X-Amzn-Requestid: 1b4a76cc-ee2e-11e6-867d-2311ebaffd3e
2017/02/08 13:40:47 [DEBUG] plugin: terraform:
2017/02/08 13:40:47 [DEBUG] plugin: terraform: <ErrorResponse xmlns="http://rds.amazonaws.com/doc/2014-10-31/">
2017/02/08 13:40:47 [DEBUG] plugin: terraform: <Error>
2017/02/08 13:40:47 [DEBUG] plugin: terraform: <Type>Sender</Type>
2017/02/08 13:40:47 [DEBUG] plugin: terraform: <Code>InvalidDBClusterStateFault</Code>
2017/02/08 13:40:47 [DEBUG] plugin: terraform: <Message>Cluster cannot be deleted, it still contains DB instances in non-deleting state.</Message>
2017/02/08 13:40:47 [DEBUG] plugin: terraform: </Error>
2017/02/08 13:40:47 [DEBUG] plugin: terraform: <RequestId>1b4a76cc-ee2e-11e6-867d-2311ebaffd3e</RequestId>
2017/02/08 13:40:47 [DEBUG] plugin: terraform: </ErrorResponse>
2017/02/08 13:40:47 [DEBUG] plugin: terraform:
2017/02/08 13:40:47 [DEBUG] plugin: terraform: -----------------------------------------------------
```
Error returns now, as expected:
```
Error applying plan:
2017/02/08 13:40:47 [DEBUG] plugin: waiting for all plugin processes to complete...
1 error(s) occurred:
* aws_rds_cluster.jake (destroy): 1 error(s) occurred:
2017/02/08 13:40:47 [DEBUG] plugin: terraform: aws-provider (internal) 2017/02/08 13:40:47 [DEBUG] plugin: waiting for all plugin processes to complete...
* aws_rds_cluster.jake: RDS Cluster cannot be deleted: Cluster cannot be deleted, it still contains DB instances in non-deleting state.
```
Our DNS tests were using terraform.test as a DNS name, which GCP was
erroring on, as we haven't proven we own the domain (and can't, as we
don't). To solve this, I updated the tests to use hashicorptest.com,
which we _do_ own, and which we have proven ownership of. The tests now
pass.
Use consul locks to implement state locking. The lock path is state path
+ "/.lock" which matches the consul cli default for locks. Lockinfo is
stored at path + "/.lockinfo".
James Bardin