stress-ng — Reporting

Collect-StressNgResults.ps1 retrieves the raw YAML output written by stress-ng on each target node, parses the per-stressor metrics, and writes normalised aggregate JSON to the logs\stress-ng\<RunId>\ directory.

---

Collection Steps

.\scripts\Collect-StressNgResults.ps1 -RunId "stressng-cpu-202412011430"

The script performs the following in sequence:

1. SCP retrieval — Copies /tmp/stress-ng-results/<RunId>/stress-ng-results.yml from each target node to logs\stress-ng\<RunId>\<node>-stress-ng-results.yml
2. YAML parse — Loads the YAML using ConvertFrom-Yaml; accesses metrics via $stressData.'stress-ng'.metrics (note: the hyphen in the key requires single-quote wrapping in PowerShell)
3. Per-stressor normalisation — Extracts bogo-ops, bogo-ops-per-second-usr-sys-time, real-time, usr-time, and sys-time for every stressor in every node's output
4. Aggregate computation — Calculates mean bogo-ops/sec and total bogo-ops across all nodes per stressor
5. Write output — Saves <RunId>-aggregate.json and <RunId>-per-node.json to logs\stress-ng\<RunId>\

---

Output Files

File	Content
logs\stress-ng\<RunId>\<RunId>-aggregate.json	Cross-node aggregate metrics per stressor
logs\stress-ng\<RunId>\<RunId>-per-node.json	Per-node, per-stressor raw metrics
logs\stress-ng\<RunId>\<node>-stress-ng-results.yml	Raw YAML from stress-ng on each node (preserved for audit)

---

Aggregate JSON Schema

{
  "run_id": "stressng-cpu-202412011430",
  "profile": "cpu-stress",
  "node_count": 4,
  "stressor_results": [
    {
      "stressor": "cpu",
      "avg_bogo_ops_per_sec": 185432.6,
      "total_bogo_ops": 3708652,
      "node_count": 4
    },
    {
      "stressor": "cpu-cache",
      "avg_bogo_ops_per_sec": 94211.0,
      "total_bogo_ops": 1884220,
      "node_count": 4
    }
  ],
  "collected_at": "2024-12-01T14:45:00Z"
}

Aggregate Field Reference

Field	Type	Description
run_id	string	Matches the -RunId parameter
profile	string	Profile name from the YAML definition
node_count	int	Number of nodes that produced results
stressor_results[].stressor	string	stress-ng stressor name (cpu, vm, hdd, etc.)
stressor_results[].avg_bogo_ops_per_sec	float	Mean bogo-ops/sec across all nodes for this stressor
stressor_results[].total_bogo_ops	int	Sum of bogo-ops across all nodes
stressor_results[].node_count	int	Nodes that reported this stressor (may be < total if OOM killed)
collected_at	ISO 8601	UTC timestamp when collection completed

---

Per-Node JSON Schema

{
  "run_id": "stressng-cpu-202412011430",
  "nodes": [
    {
      "node": "hci01-node1",
      "stressors": [
        {
          "stressor": "cpu",
          "bogo_ops": 927163,
          "bogo_ops_per_sec_usr_sys": 185432.6,
          "real_time_s": 300.1,
          "usr_time_s": 4800.4,
          "sys_time_s": 0.3
        }
      ]
    }
  ]
}

---

YAML Key Note

The raw stress-ng output nests metrics under the stress-ng key, which contains a hyphen. In PowerShell, accessing this key requires single-quote notation:

$raw     = Get-Content "hci01-node1-stress-ng-results.yml" -Raw
$parsed  = $raw | ConvertFrom-Yaml
$metrics = $parsed.'stress-ng'.metrics   # Must quote the hyphenated key

---

Interpreting Results

Bogo-ops vs. Bogo-ops/sec

• total_bogo_ops is the raw operation count over the full timeout window. Use this for absolute throughput comparisons across runs of equal duration.
• avg_bogo_ops_per_sec normalises for wall-clock time. Use this when comparing results where one node throttled and finished fewer operations than expected.

Identifying Outlier Nodes

$perNode = Get-Content "logs\stress-ng\stressng-cpu-202412011430\*-per-node.json" |
    ConvertFrom-Json

$perNode.nodes | ForEach-Object {
    $node = $_.node
    $_.stressors | Where-Object { $_.stressor -eq "cpu" } |
        Select-Object @{n="Node";e={$node}}, bogo_ops_per_sec_usr_sys
} | Sort-Object bogo_ops_per_sec_usr_sys

A node reporting bogo-ops/sec more than 15% below the cluster mean likely experienced throttling or interference. Review monitoring logs for stressng_cpu_throttling alert events recorded during that run window.

---

Report Generation

Aggregate results are available as input for the reporting template:

.\scripts\New-LoadReport.ps1 -RunId "stressng-cpu-202412011430" -Tool stress-ng `
    -TemplatePath "reports\templates\stress-ng-report.adoc" `
    -OutputPath "reports\stressng-cpu-202412011430.pdf"

The template uses the following aggregate JSON fields as placeholders:

Placeholder	Source Field
{run_id}	run_id
{profile}	profile
{node_count}	node_count
{cpu_avg_bogo_ops_per_sec}	stressor_results[stressor=cpu].avg_bogo_ops_per_sec
{memory_avg_bogo_ops_per_sec}	stressor_results[stressor=vm].avg_bogo_ops_per_sec
{io_avg_bogo_ops_per_sec}	stressor_results[stressor=hdd].avg_bogo_ops_per_sec
{collected_at}	collected_at