Question

我们一直使用以下方法在远程计算机上担任角色一个小时以上：

# Prep environment to use roles.
unset AWS_CONFIG_FILE
unset AWS_DEFAULT_REGION
unset AWS_DEFAULT_PROFILE

CONFIG_FILE=$(mktemp)

# Creates temp file with instance profile credentials as default
# AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, ROLE_ARN are available from the environment.
printf "[default]\naws_access_key_id=$AWS_ACCESS_KEY_ID\naws_secret_access_key=$AWS_SECRET_ACCESS_KEY\n[profile role_profile]\nrole_arn = $ROLE_ARN\nsource_profile = default" > $CONFIG_FILE

# make sure instance profile takes precedence
unset AWS_ACCESS_KEY_ID
unset AWS_SECRET_ACCESS_KEY
unset AWS_SESSION_TOKEN

export AWS_CONFIG_FILE=$CONFIG_FILE
export AWS_DEFAULT_REGION=us-east-1
export AWS_DEFAULT_PROFILE=role_profile

不幸的是，该方法最近开始失败。我们可以通过运行以下命令来重现故障：

aws sts get-caller-identity

在最后一个命令中添加--debug标志：

09:11:47 2018-06-21 14:11:47,731 - MainThread - awscli.clidriver - DEBUG - CLI version: aws-cli/1.15.40 Python/2.7.12 Linux/4.9.76-3.78.amzn1.x86_64 botocore/1.10.40
...
09:11:47 2018-06-21 14:11:47,811 - MainThread - botocore.hooks - DEBUG - Event choose-signer.sts.GetCallerIdentity: calling handler <function set_operation_specific_signer at 0x7f22d19a6ed8>
09:11:47 2018-06-21 14:11:47,812 - MainThread - botocore.credentials - WARNING - Refreshing temporary credentials failed during mandatory refresh period.
09:11:47 Traceback (most recent call last):
09:11:47   File "/var/lib/jenkins/.local/lib/python2.7/site-packages/botocore/credentials.py", line 432, in _protected_refresh
...
09:11:47     raise KeyError(cache_key)
09:11:47 KeyError: 'xxxx' (redacted)
09:11:47 2018-06-21 14:11:47,814 - MainThread - awscli.clidriver - DEBUG - Exiting with rc 255

显然，Python“缓存”字典中缺少键。

Answer 1

显而易见的解决方案是找到并删除缓存：

rm ~/.aws/cli/cache

但是，这并不能解释它是如何开始发生的（以及是否会再次发生）。谁能解释发生了什么事？

Answer 2

~/.aws/cli中的权限可能是错误的。

检查权限：

ls -la ~/.aws/cli
ls -la ~/.aws/cli/cache

您的文件可能具有错误的权限或所有权。更正它们，aws cli命令将正常工作。

~/.aws/cli/cache中的文件所需的权限为-rw-------。

希望有帮助。

无法通过AWS配置文件承担角色

2 个答案: