Managing Amazon Msk

Amazon MSK

Overview

Domain expertise for operating Amazon MSK Provisioned clusters with Standard and Express broker types. Covers performance troubleshooting, consumer lag diagnosis, storage management, cluster sizing, client configuration, and CloudWatch monitoring.

Execute commands using available tools from the AWS MCP server when connected — it provides sandboxed execution, audit logging, and observability. When the MCP server is not available, fall back to the AWS CLI or shell as needed.

Standard brokers use customer-managed EBS volumes for storage. You choose instance types (kafka.m5/m7g families), provision EBS, and manage storage scaling.

Express brokers provide fully managed, pay-as-you-go storage with no EBS provisioning. They use instance types prefixed with express.m7g, offer up to 3x more throughput per broker, and have no maintenance windows. Express brokers have NO customer-managed EBS — do NOT recommend EBS expansion or provisioned throughput for Express clusters. Express brokers enforce fixed replication factor of 3 and min.insync.replicas=2 — do NOT attempt to create topics with RF=1 on Express. If RF=1 is needed, use Standard brokers.

Which Workflow Do You Need?

Determine the broker type first: aws kafka describe-cluster-v2 --cluster-arn <arn>. Check Provisioned.BrokerNodeGroupInfo.InstanceType — if it starts with express., it is an Express cluster.

Available scripts

• scripts/msk_sizing.py — MUST be run for any sizing question (broker count, instance choice, cost). See size-and-choose-cluster.md (opens in a new tab) for the required workflow and script reference.

Guardrail — where this skill’s own files live (MCP vs local install)

This skill can be loaded two ways, and they resolve the skill’s own bundled files — the references/ documents and the scripts/ files from different places. Determine how the skill was loaded before you read a reference or run a script:

• Loaded through the AWS MCP retrieve_skill tool call. The skill is not installed on the local filesystem; its reference files and scripts do not exist on disk. You MUST fetch each reference or script through the same retrieve_skill tool by passing the file parameter (for example, file="references/configure-clients.md" or file="scripts/msk_sizing.py"), and run a script from the content that tool returns. Do NOT file_read these paths from the local or working directory, and do NOT search the filesystem for them — they are not there, and any local file that happens to match the name is unrelated to this skill.
• Installed locally (the skill lives in a local skills directory such as .claude/skills/managing-amazon-msk/, ~/.claude/skills/managing-amazon-msk/, or .kiro/skills/managing-amazon-msk/). Read references and run scripts from the local skill directory using the relative paths shown throughout this documentation.

This distinction applies only to the skill’s own packaged files. Every artifact created during a session or supplied by users are read from and written to the user’s working directory regardless of how the skill was loaded. Never fetch or write customer data through retrieve_skill.

Common Workflows

Create cluster configuration (server.properties):

The --server-properties argument MUST be a real Kafka properties file with one key=value per line, separated by actual newline (\n) characters — NOT the literal two-character escape sequence \n. The MSK API accepts the bytes as-is; if you pass "k1=v1\nk2=v2" as a single string with escaped newlines, MSK stores ONE invalid property line and the cluster will fail to apply it.

Recommended pattern: write the properties to a local file with real newlines, then pass it via fileb:// so the CLI uploads the raw bytes verbatim. Verify by reading the revision back with describe-configuration-revision and base64-decoding ServerProperties — you should see one property per line.

cat > server.properties <<'EOF'
auto.create.topics.enable=false
default.replication.factor=3
min.insync.replicas=2
unclean.leader.election.enable=false
num.io.threads=32
num.network.threads=16
log.retention.hours=168
EOF

aws kafka create-configuration \
  --name <config-name> \
  --kafka-versions "3.6.0" \
  --server-properties fileb://server.properties

For per-instance-size thread tuning (num.io.threads, num.network.threads) and durability defaults, see size-and-choose-cluster.md (opens in a new tab) and configure-clients.md (opens in a new tab).

Additional Resources

• MSK Best Practices - Standard (opens in a new tab)
• MSK Best Practices - Express (opens in a new tab)
• MSK Client Best Practices (opens in a new tab)
• MSK CloudWatch Metrics (opens in a new tab)
• MSK Quotas (opens in a new tab)
• MSK Configuration