Troubleshooting¶

This page lists the common failures, what causes them and what to do. Drive quirks are described in full on each drive's page under Drives; this page gives the short version and links there.

When something fails and the message does not say why, run the command again with -v before the subcommand (tw -v drive status). Debug logging shows every decision, guard and command, and which config file and profile were picked up.

The board does not answer¶

The drive does not answer¶

tw stops with no drive answered auto-detection (tried: ...). None of the wake-ups auto tried got an answer. Work down this list:

1. Power. Check that the drive gets both 5 V and 12 V from its own supply.
2. Cable. Check that the cable is pin 1 to pin 1 at both ends.
3. Profile. Read the auto: log lines. A drive that needs a profile auto never tries (teac, conner, iomega, or your own file) needs --profile NAME. auto: skipping insight: ... means the board refused the motor lines; the other wake-ups were still tried.
4. Config file. A profile in the config file replaces auto. tw -v drive status logs which file and profile were loaded.
5. Unit address. Try another phantom unit address: tw --profile colorado drive select --unit 1.
6. Another drive is selected. A phantom drive left selected by an earlier session can hold the bus. Run tw drive deselect, or switch the drives' power off and on.
7. Look at the lines. tw --profile colorado drive scope logs edges on TRK0 and INDEX. A selected, ready drive shows cue INDEX pulses every few milliseconds; no edges at all points at power, cabling or selection.

If a named profile (--profile colorado, mountain or insight) works where auto does not, auto is at fault. Only the Colorado wake-up has been tested on real drives. Name the profile and report the problem.

The drive goes dark (TEAC FT-3010)¶

The TEAC stops answering on every line, and only a power cycle brings it back. This happens when a report command is interrupted part-way, including an ACK time-out or tw drive scope --pulses N. Use --profile teac, never auto, and do not send it stray pulses. See TEAC FT-3010.

The dump will not start¶

The Colorado 1400 will not read a QIC-80 tape¶

The bench Colorado 1400 never references a QIC-80 tape, and refuses tw drive rate 500 and tw drive format qic80 with error 31 (Rate or Format Selection Error). Dump QIC-80 tapes on a QIC-80 drive such as the Jumbo 350. See Colorado 1400.

The dump stops part-way¶

tw dump prints dump stopped: track N: REASON. After every pass it checks that the pass ended at the end of the tape, that it found at least 90% as many segments as the best pass so far, and that it missed no more than 5% of its segments. With --check it also stops when fewer than 80% of the sectors are CRC-clean. Any of these usually means a dirty head or a failing tape.

1. Read the reason.
2. Clean the head (A dirty head).
3. Resume in place: run the same command with --no-clobber. Tracks already dumped clean are skipped; the track that stopped is read again.
4. Add --retries 2 so a flaky pass is read again before the health check judges it.

No header found¶

qicsilver identify says no header segment found: the capture must include the start of track 0, or neither copy of the header segment was recovered.

• Make sure the image includes track 0: tw dump captures/mytape 0, then convert again.
• Check the rate in drive config (tw drive status). A capture at the wrong rate decodes no sectors.
• Dump track 0 again into a new directory and convert both directories together.

• Record a short sample and count what decodes:

  tw --profile colorado drive flux --motion logical --seconds 10
  

  Zero sectors means a head, rate or tape problem, not a software one. Use --motion logical: the default --motion fwd runs the tape at high speed, where sectors never decode. See tw drive.

Wrong volume profile¶

The volume table reads as nonsense: dates in the wrong century, absurd sizes, or extract failing on the first volume. The layout of a volume-table entry depends on the backup program that wrote it, and qicsilver guessed wrong.

1. Run qicsilver identify --raw captures/mytape.twtz to see each profile's score.
2. Force the right one with --volume-profile NAME on identify, extract, inspect and tar.
3. If none fits, set the layout by hand with the override flags.

See Cartridge and volume profiles and Manual read overrides.

A dirty head¶

Old tapes shed oxide onto the head. Signs:

• A dump that stops with fewer segments found than on earlier tracks.
• --check failing below 80%.
• A track that converts much worse than its neighbours.

Stop. Clean the head and the capstan with isopropyl alcohol and a lint-free swab, and let them dry. Then resume the dump with --no-clobber, or dump the remaining tracks into a new directory and convert the directories together.

tw drive clean runs a cleaning cartridge back and forth. It is untested on the bench: the cleaning cartridge on hand has never tripped a test drive's cartridge switch. See Cartridges.

Segments are uncorrectable after convert¶

tw convert reports uncorrectable segments: those had more bad sectors than the error correction can repair (3 per segment).

• Dump the affected tracks again into a new directory and convert every directory together. A sector read on any pass fills the gaps of the others.
• Keep every read with tw dump --keep-all or --passes N. With three or more reads of a sector, tw convert majority-votes it byte by byte.
• Look at the damage in cinnabar to see which tracks and segments are worth reading again.

See tw convert.

See also¶

• tw drive
• Configuration and logging
• Drives