Admin Console β Error Log Viewer
The Error Log Viewer in the SmartBot admin console lets super admins query aggregated error patterns across SmartBot's processing pipeline without needing access to the underlying cloud infrastructure. Use it to identify recurring issues, investigate specific error occurrences, and monitor the health of your SmartBot deployment.
Audience: Super admins responsible for monitoring and troubleshooting SmartBot behavior across the organization.
Before you begin
- You must have super admin access to the SmartBot admin console. Standard admin accounts do not have access to the Error Log Viewer.
- Error log queries are asynchronous β queries are submitted and results are retrieved separately. Allow time for the query to complete before expecting results.
- Queries span all SmartBot processing components, including request handling, response delivery, analytics, webhooks, WebSocket connections, and batch jobs.
Steps
Query error patterns for a time range
Use this procedure to identify recurring error patterns across your SmartBot deployment during a specific window of time.
- Sign in to the SmartBot admin console.
- Navigate to the Error Log Viewer. This section is accessible to super admins only.
- Specify the time range for your query. You can express the range as a relative window (for example, "last 2 hours") or as a specific start and end time.
- Submit the query. The system accepts your query immediately and begins processing in the background.
- Note the query ID returned in the confirmation. You will use this ID to retrieve results when the query completes.
- Poll for results using the query ID. Repeat until the query status indicates completion.
Once complete, the results include:
- Error patterns β Each distinct error pattern found, along with:
- Pattern text describing the error
- Sample count (how many times the pattern appeared)
- Percentage of total errors that this pattern represents
- A time histogram showing when the pattern occurred within your selected window
- Up to 3 sample log lines for each pattern
- Aggregate statistics β Summary counts for the query as a whole:
- Records matched
- Records scanned
- Bytes scanned
Note: The percentage shown for each error pattern is calculated against the total number of error records matched within the query window, not against total log volume.
Investigate a specific error occurrence with surrounding context
When you have identified a specific error occurrence β for example, from a guest-reported issue or from an error pattern's sample log lines β use this procedure to retrieve the log lines immediately before and after that event.
- Locate the request ID and timestamp for the error you want to investigate. You can find these in the sample log lines returned by an error pattern query.
- Open the surrounding logs lookup in the Error Log Viewer.
- Enter the request ID and the timestamp of the specific error occurrence.
- Submit the lookup. The system retrieves the log lines recorded immediately before and after the specified event.
- Review the surrounding log lines to understand the context of the error β for example, what request preceded it, which component handled it, and what followed.
Tip: Use surrounding log context to distinguish between isolated failures and systemic issues. If the surrounding lines show normal processing before and after, the error may be a transient exception rather than an ongoing problem.
Important: The surrounding logs lookup requires both a valid request ID and a timestamp. Queries submitted with an unrecognized request ID will return no results. Obtain both values from a prior error pattern query or from the SmartBot inbox conversation view where available.
Tips
- Start with a broad time window, then narrow. Begin with a longer range (such as "last 24 hours") to establish a baseline of error patterns, then re-query a shorter window around the time a specific issue was reported to isolate its causes.
- Use sample log lines to identify request IDs. Each error pattern includes up to 3 sample log lines. These lines typically contain the request ID you need to perform a surrounding logs lookup for a specific occurrence.
- Check percentage of total errors, not only sample count. A pattern with a high sample count but a low percentage may be less significant than one with a moderate count but a high percentage, especially if total error volume is low.
- Monitor the time histogram for spike patterns. If a pattern's histogram shows errors concentrated in a narrow time band, correlate that window with deployment changes, high-traffic periods, or guest-reported incidents during that time.
- Poll at reasonable intervals. Because queries run asynchronously, wait a few seconds between poll attempts. Very large time windows or high error volumes may require additional processing time.
Troubleshooting
The query returns no results for a time range when errors are expected.
Verify that the time range specified covers the period when errors occurred. If you used a relative window ("last 2 hours"), confirm you are polling within the same session β the query window is fixed at submission time. If errors occurred outside the window, resubmit with an adjusted range.
The surrounding logs lookup returns no results.
Both the request ID and the timestamp must be valid and must correspond to a log event that falls within a retained log window. If the event is older than the retention period for SmartBot logs, surrounding context will not be available. Recheck the request ID and timestamp values from the sample log lines for accuracy.
A query ID is returned but results never appear when polling.
Very broad time windows or high error volumes can cause queries to take longer than expected. Continue polling at intervals. If the query remains in a pending state for an extended period without completing, resubmit the query with a narrower time range to reduce the scope.
Error pattern percentages do not add up to 100%.
Each pattern's percentage is calculated against total matched records for the query. Patterns with very low occurrence counts may be rounded or omitted from the summary. The aggregate statistics section shows the total records matched, which you can use to verify the figures manually.
Sample log lines show truncated messages.
Log lines are returned as stored. If the original log message exceeded the line length limit at the time it was recorded, the line will appear truncated. Use the surrounding logs lookup with the request ID from the truncated line to retrieve additional context.