Ubuntu服务器CodeBuild项目RDS克隆命令格式错误排查求助

Let's break down the issues with your loop command and surrounding configuration that's causing your CodeBuild task to fail:

1. Undefined Variables ($DBNAME and $DBIDENTIFIER) #

Looking at your pre_build phase, you define $TARGETDBENV as your new cluster name, but later reference $DBNAME and $DBIDENTIFIER in the loop and endpoint retrieval commands. These variables are not set anywhere in your script, so the AWS CLI commands will fail because they're trying to describe a cluster with an empty identifier.

2. Unnecessary HTML Entities (") in Shell Commands #

In your original loop command (and some other script sections), you're using "—an HTML entity for double quotes. Shell commands don't recognize these; you should use regular double quotes (") instead.

3. Brittle JSON Parsing with grep/awk #

Parsing AWS CLI output with grep and awk is error-prone, as JSON formatting can vary slightly. The AWS CLI has a built-in --query parameter that lets you extract values directly without relying on text manipulation tools, which is far more reliable.

Corrected pre_build Phase Commands #

Here's the fixed version of your pre_build phase with all issues addressed:

pre_build:
  commands:
    - pip install --upgrade pip
    - pip3 install awscli --upgrade --user
    - export SOURCEDBENV=mysql-dev
    - export DATE=`date +%Y%m%d`
    - export TARGETDBENV=$SOURCEDBENV-$DATE
    - echo $TARGETDBENV
    # Get latest snapshot ARN directly with AWS CLI query (no need for tr later)
    - export ARNSNAP=$(aws rds describe-db-cluster-snapshots --db-cluster-identifier $SOURCEDBENV --query "reverse(sort_by(DBClusterSnapshots, &SnapshotCreateTime))[0].DBClusterSnapshotArn" --output text)
    - echo $ARNSNAP
    - aws rds restore-db-cluster-from-snapshot --snapshot-identifier $ARNSNAP --db-cluster-identifier $TARGETDBENV --engine aurora-mysql
    - aws rds create-db-instance --db-instance-identifier $TARGETDBENV --db-instance-class db.t3.medium --db-subnet-group-name db_subnet_grp_2019 --engine aurora-mysql --db-cluster-identifier $TARGETDBENV
    # Fixed loop using correct variable and reliable status check via AWS CLI query
    - while [ $(aws rds describe-db-clusters --db-cluster-identifier $TARGETDBENV --query "DBClusters[0].Status" --output text) != "available" ]; do echo "sleep 60s"; sleep 60; done
    - echo "Temp db ready"
    # Get endpoint cleanly with AWS CLI query instead of grep/awk
    - export ENDPOINT=$(aws rds describe-db-cluster-endpoints --db-cluster-identifier $TARGETDBENV --query "DBClusterEndpoints[?contains(Endpoint, '-ro-') == false].Endpoint" --output text)
    - echo $ENDPOINT

Key Improvements Explained #

• Variable Consistency: We use $TARGETDBENV everywhere instead of undefined variables, ensuring all commands reference the correct cluster name.
• HTML Entity Removal: Replaced " with standard double quotes so the shell can interpret commands correctly.
• Reliable Status Check: The loop now uses aws rds describe-db-clusters with a --query parameter to directly fetch the cluster status, avoiding grep counts that can break from unexpected JSON formatting.
• Clean Endpoint Retrieval: Endpoint extraction uses a query to filter out read-only endpoints and return the exact string, eliminating fragile grep/awk chains.
• Simplified Snapshot ARN Fetch: Added --output text to the snapshot ARN command, so we don't need tr -d '"' to strip quotes later.

After making these changes, your CodeBuild task should successfully wait for the RDS cluster to become available before proceeding with build steps.

内容的提问来源于stack exchange，提问作者John Fox