thadam inspect

Read-only views of a data directory: WAL, cursors, segments, samples

Usage: thadam inspect <COMMAND>

Commands:
  wal       Show the WAL's segments, durable end, epoch, and retained bytes
  cursors   Show each durable cursor's committed position and in-flight range
  segments  Show a sink's file counts and bytes per catalog state, newest files first
  sample    Print the raw payloads of the newest frames, truncated. Prints at most the first 5 records of each frame
  help      Print this message or the help of the given subcommand(s)

Options:
  -h, --help  Print help

thadam inspect reads a data directory without changing it. It is safe to run while thadam run holds the data directory: it never takes the data-directory lock or the WAL’s writer lock, and it never repairs, truncates, or rewrites a file. Each subcommand prints human-readable text by default and JSON with --json.

Exit codes: 0 on success, 2 for a usage error, 3 when the data directory is not a readable WAL (see exit codes), 1 when a file cannot be read.

thadam inspect wal

Show the WAL's segments, durable end, epoch, and retained bytes

Usage: thadam inspect wal [OPTIONS] <DATA_DIR>

Arguments:
  <DATA_DIR>  The data directory (`pipeline.data_dir` in the configuration)

Options:
      --json  Print JSON instead of text
  -h, --help  Print help

Prints the WAL epoch, node and machine identity, whether the last shutdown was clean, the first and next positions, the durable end (the last position that survived an fsync), retained bytes, torn and zero tails, and each segment file.

The --json output has these keys: dir, epoch, node_id, machine_id, clean_shutdown, first_seq, next_seq, durable_end, retained_bytes, torn_tail, zero_tail, segments (each with base, path, bytes), and cursor_count.

thadam inspect cursors

Show each durable cursor's committed position and in-flight range

Usage: thadam inspect cursors [OPTIONS] <DATA_DIR>

Arguments:
  <DATA_DIR>  The data directory (`pipeline.data_dir` in the configuration)

Options:
      --json  Print JSON instead of text
  -h, --help  Print help

One entry per durable cursor. A sink’s cursor is named sink.<id>. Each entry shows the committed position and, when a batch is being processed, the in-flight range (inflight is null otherwise).

The --json output is {"cursors": [{"name", "sink", "committed", "inflight"}]}, where inflight is {"first", "last", "frames"} or null.

thadam inspect segments

Show a sink's file counts and bytes per catalog state, newest files first

Usage: thadam inspect segments [OPTIONS] --sink <ID> <DATA_DIR>

Arguments:
  <DATA_DIR>  The data directory (`pipeline.data_dir` in the configuration)

Options:
      --sink <ID>  The sink whose catalog to read
      --json       Print JSON instead of text
  -h, --help       Print help

Reads the sink’s segment catalog (sinks/<id>/catalog.log) and prints the file count and bytes for each state: finalized, eligible, uploaded, and deleted, followed by the newest files with their dataset, status, position range, row count, and bytes.

The --json output is {"sink", "totals", "files"}, where totals maps each state to {"files", "bytes"} and each entry in files has id, dataset, status, path, first, last, rows, and bytes.

thadam inspect sample

Print the raw payloads of the newest frames, truncated. Prints at most the first 5 records of each frame

Usage: thadam inspect sample [OPTIONS] <DATA_DIR>

Arguments:
  <DATA_DIR>  The data directory (`pipeline.data_dir` in the configuration)

Options:
  -n, --count <N>  How many of the newest frames to print [default: 5]
      --json       Print JSON instead of text
  -h, --help       Print help

Prints the raw payloads of the newest retained frames, before any chain runs. Each payload is cut at 200 bytes on a character boundary and marked with ...; at most 5 records are printed per frame.

The --json output is {"frames": [{"seq", "source", "clock_synced", "record_count", "records"}]}, where each record has ingest_time_ns, payload_bytes, payload, and truncated.