Platform observability and SLAs in MLOps - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to define a basic metric for monitoring model latency.

MLOps

model_latency = [1]('model_latency_seconds')

Drag options to blanks, or click blank then click option'

AHistogram

BGauge

CCounter

DSummary

Attempts:

3 left

💡 Hint

Common Mistakes

Using Counter which only counts events, not durations.

A Histogram is used to observe the distribution of latency values over time, which is ideal for model latency monitoring.

2fill in blank

medium

Complete the code to set an alert threshold for SLA violation when error rate exceeds 5%.

MLOps

if error_rate [1] 0.05:
    trigger_alert()

Drag options to blanks, or click blank then click option'

A<=

B<

C>

D==

Attempts:

3 left

💡 Hint

Common Mistakes

Using '<' which triggers alert for low error rates.

The alert triggers when the error_rate is greater than 5% (0.05), indicating SLA violation.

3fill in blank

hard

Fix the error in the Prometheus query to calculate 99th percentile latency.

MLOps

histogram_quantile(0.99, sum(rate([1][5m])) by (le))

Drag options to blanks, or click blank then click option'

Amodel_latency_seconds_sum

Bmodel_latency_seconds_bucket

Cmodel_latency_seconds_count

Dmodel_latency_seconds

Attempts:

3 left

💡 Hint

Common Mistakes

Using raw latency or count metrics instead of buckets.

The histogram_quantile function requires the bucketed histogram metric, which ends with _bucket.

4fill in blank

hard

Fill both blanks to create a dictionary comprehension that maps model names to their error rates above 1%.

MLOps

{model: [1] for model, [2] in metrics.items() if error_rate > 0.01}

Drag options to blanks, or click blank then click option'

Aerror_rate

Clatency

Dcount

Attempts:

3 left

💡 Hint

Common Mistakes

Using latency or count instead of error_rate.

We want to map each model to its error_rate when the error rate is above 1%. Both blanks use error_rate.

5fill in blank

hard

Fill all three blanks to filter logs for errors, extract timestamps, and count occurrences.

MLOps

error_counts = {log['[1]']: logs.count([2]) for log in logs if '[3]' in log['message']}

Drag options to blanks, or click blank then click option'

Atimestamp

Blog

Cerror

Dlevel

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'level' instead of 'error' for filtering.

The dictionary keys are timestamps, the count is of the log itself, and the filter checks if 'error' is in the message.