Error stacks and runner diagnostics

A run's row in a log keeps more than its status and duration. A failed run keeps the full text of its error, and a Connect Any API load keeps a report of what it did. Both open from the row, in the same window: it is titled Error Stack Trace when the run kept a stack, and Runner Diagnostics when it kept none.

You meet the window in a table's run log: the Sync Log of a staging table or a Writeback table, or the Run Log on a transformation's Execution tab. For the shared table, see Sync Log tables; for the logs themselves, see Run, stop and troubleshoot loads, Synchronize a table and read the Sync Log and Deploy, run, truncate and monitor tables.

Open a failed run's error stack

1. Find the failed run in the log. Its error cell previews the first line of the message (the column is Error, or Error Stack in a staging table's Sync Log).
2. Click the cell to open the window. The full message appears in a read-only view, and a long stack scrolls within the window.
3. For a Connect Any API load, the loader's report sits above the raw stack, so start there when the stack alone does not explain the failure.
4. Click Copy to take the details with you; the label turns Copied for a moment. Close dismisses the window, and so does the X in the header.

Open a load's runner diagnostics

A Connect Any API load that finished with warnings keeps an amber badge in its error cell. The badge reads UTF-8 replaced: 4 when the count of replaced sequences is known, or 3 warnings (or 1 warning) when it is not.

1. Find the run with the amber badge in its error cell.
2. Click the badge to open the window. There is no stack to show, so it opens as Runner Diagnostics, with a warning icon and amber tones instead of the red failure ones. Screen readers name the badge's control View runner diagnostics.
3. Read the report, as the next section describes. Copy takes it with you, and Close dismisses the window.

A badge opens the window only when the row keeps a report; without one it stays a plain label, showing its summary on hover.

Read the loader's report

The report describes one load attempt, and it holds only what the loader recorded.

• The headline badge names the outcome: Load failed or Load finished with warnings.
• Beside it sit the loader's identifiers and totals: Stream, Mode, Pages, Rows, Elapsed, Parallel, Runner and CDK. Elapsed reads in seconds to one decimal, like 12.3 s.
• The loader's message, when there is one, appears below in its original shape, including line breaks.
• A Page section describes the page request: Params, Index, Token and URL.
• An HTTP section shows what came back: Status, Content-Type and Bytes.
• A Decode section appears when a response could not be read: Kind, the Error, the position of the problem and the Bad bytes in hex, like 0xED 0xA0 0x80. The position reads Byte for Invalid UTF-8 and Char for the other kinds, and a short block shows the surrounding text when the loader kept it.
• A Warnings section lists one card per warning: its sentence, like 4 invalid UTF-8 sequences replaced in 2 records; the First at byte and Bytes of the first occurrence; its Page and Pages; and the affected records, each line ending with -> field. A list that was cut short ends with ... more records not listed, and record keys you may not read show as [hidden: no data access].
• A Log tail (N) fold holds the loader's last log lines. It starts folded shut; click it to open, and click again to fold.

Fields and sections without content are left out rather than shown empty. The four Kind values of the Decode section:

For the messages a failed Connect Any API load can show and what to do about them, see Lake troubleshooting.

Questions, answered

Why does the window say Runner Diagnostics instead of Error Stack Trace?

The title follows what the row holds. A stack opens as Error Stack Trace; a report with no stack opens as Runner Diagnostics. The window is the same either way.

Why is the window amber and not red?

Amber marks the warning-only window: no stack, just the report of a run that finished with warnings, with a warning icon in the header. A window that carries a stack, or a failed load, keeps the red failure tones.

The run finished, so why is there a warning on its row?

The load met invalid UTF-8 sequences in the source and replaced them so the run could finish, and the badge counts what it replaced. Open the row's report to read the exact sentence, like 3 invalid UTF-8 sequences replaced in 1 record, and to see which records were affected. The data was loaded; the loader replaced those sequences to keep reading.

Why is the amber badge not opening a window?

A badge opens the window only when the row keeps a report. When it keeps none, the badge stays a plain label: hovering it shows the summary sentence, and clicking does nothing.

What do the Decode kinds mean?

Invalid UTF-8: the response held bytes that are not valid UTF-8. Invalid JSON: the response body is not valid JSON. HTML body: the response is an HTML page. Empty body: the response body is empty. The position next to them reads Byte for Invalid UTF-8 and Char for the other kinds.

Why are some fields missing from the report?

The report leaves out what the loader did not send: a field with no value is skipped, and whole sections such as Page, HTTP, Decode, Warnings and the log tail appear only when they have content. A short report is not a broken one.

What does Copy put on my clipboard?

The window's content as one text: the raw stack when there is one, and the loader's report as JSON below it. In a window that holds only the report, you get the report alone. Paste it into a message or a ticket when you need help reading it.

What does "... more records not listed" mean?

The warning found more affected records than the list shows. The list stops early, and the count in the sentence above carries a plus sign, so 10+ records means ten or more.

What is the Log tail?

The loader's last log lines for the attempt, folded shut under the report. Click Log tail (N) to open it, where N is the number of lines; click again to fold it back.

Can I reopen the window after I close it?

Yes. The stack and the report stay on the run's row, so clicking the error cell or the badge again opens the same window. Close only dismisses it and changes nothing about the run.