Assessment Sizing

1. Reads migrate-to-msk-skill-artifacts/<cluster_name>/cluster-config.json (discovery contract).
2. Computes the six workbook input values (see mapping below).
3. Writes msk-sizing-inputs.<cluster_name>.json — a small JSON artifact that records each value, the cell it maps to, and the workbook source page.
4. When given --workbook <path> (a workbook the agent downloaded), fills the six input cells on the MSK Provisioned sheet and writes the filled MSK_Sizing_Pricing.<cluster_name>.xlsx. Without --workbook, it prints a cell-by-cell fill-in table instead.

No sizing math, no pillar verdict, no spreadsheet library, and no network access — the script fills a workbook the agent has already downloaded, using only the Python standard library (zipfile + re).

Filling the workbook (agent flow)

1. Resolve the workbook download from the AWS page. Read the Express best-practices page (opens in a new tab) (section “Right-size your cluster”) and follow its MSK Sizing/Pricing workbook hyperlink to download the .xlsx into the working directory. Do not hardcode a download URL — resolve it from the page each time so the skill always uses the current AWS-published workbook.
2. Fill it. Run sizing.py with --workbook <downloaded.xlsx>. The script computes the six inputs from the discovery contract, writes them into the MSK Provisioned sheet, and saves migrate-to-msk-skill-artifacts/<cluster_name>/MSK_Sizing_Pricing.<cluster_name>.xlsx. It also writes msk-sizing-inputs.<cluster_name>.json recording the cell-to-value mapping. The script never downloads anything itself.
3. Read the recommendations. Open the filled workbook in Excel, LibreOffice, or Google Sheets. The formulas recalculate on open (the script sets fullCalcOnLoad); then read the recommended instance type, broker count, and monthly cost (see “Reading the recommendations”).

If the workbook download is unavailable (for example, no network access), run sizing.py without --workbook: it writes the JSON inputs and prints a cell-by-cell fill-in table the customer can enter into the workbook manually.

Cell mapping

Sheet: MSK Provisioned (cluster design inputs section, rows 11–21).

Cells the customer does not change (workbook defaults are used):

• C15 Utilization at Peak (Standard) — default 0.5
• C16 Utilization at Peak (Express) — formula =C15*1.5 = 0.75
• C18 Retention in primary storage (tiered) — default 24
• C19 Provisioned Storage Throughput — default 1000
• C21 Replication Factor — default 3 (Express forces a replication factor of 3 anyway)
• C24 Number of AZs — default 3 (Express requires 3)
• C25 Nearest Replica Fetching — default true
• C26 EBS Disk Utilization — default 0.5
• C29–C31 EC2 comparison inputs

Caveats

1. Average throughput isn’t in the discovery contract. The workbook uses average for storage volume and cost projection (not for sizing math itself). The script defaults to peak / 2 as a rough heuristic; over-estimates cost for steady workloads and under-estimates for spiky ones. For accurate cost projection, supply real averages via --avg-in-mbps and --avg-out-mbps.

2. Retention is per-topic in the discovery contract; the workbook takes one number. The script emits the max over topics as an upper-bound storage estimate. Override via --retention-hrs if the source has a small set of low-retention topics dominating the storage picture.

3. The workbook hard-codes us-east-1 pricing. Cell I8 says us-east-1 pricing, so the cost figures do not reflect other AWS Regions.

Reading the recommendations

After opening the filled workbook in a spreadsheet app:

• Express recommendations: rows 26–32 (express.m7g.large through express.m7g.16xlarge). Instance type is in column G (G26:G32), recommended broker count in column H, and monthly cost in column I (I26:I32).
• Bottleneck breakdown: rows 134–155 show, per instance, which constraint (ingress / egress / partitions / storage) drives the broker count. The Express instances occupy cells F149:H155, the range the assessment response points customers to.

The workbook does not pick a primary; the human reviewer picks based on operational preference (fewer big brokers vs more small brokers) and cost.

Refresh procedure

When AWS publishes an updated workbook, no change to this skill is required — the workbook is downloaded fresh each time. If AWS shifts the input cells, update the CELL_* constants in sizing.py and the mapping above.

Source

The workbook download link lives on the AWS Express best practices — Right-size your cluster (opens in a new tab) page, under the “MSK Sizing/Pricing workbook” link. Resolve the download from that page rather than hardcoding a direct URL — AWS may change where the workbook is hosted.

Security considerations

• Sizing inputs reveal capacity and topology details. The msk-sizing-inputs.<cluster_name>.json artifact and the filled workbook contain peak throughput, partition count, retention, and the recommended broker count and instance type for the workload. Treat them as sensitive — these details are useful inputs for targeted attacks. Do not share via unencrypted email, public channels, or public ticketing systems without redaction.
• Store with restrictive permissions. Keep the artifact inside migrate-to-msk-skill-artifacts/<cluster_name>/ and apply restrictive permissions (e.g., chmod 600) appropriate for your environment.
• Discovery input may contain credentials. sizing.py only reads workload metrics and topic counts from cluster-config.json, but the discovery contract may also include security profile fields. Before processing, verify the input does not contain credential material (SASL passwords, API keys, private keys); redact if necessary.