## TL;DR
Treat the completed job as suspect until you read back from that cartridge: note the exact media error counts, the cartridge and drive involved, and whether errors cluster on one tape, one drive, or all media. Clean the drive, retry a read or verify of the affected cartridge on a different drive, and retire media that errors in both places.

```text
tape backup completes with warnings media errors logged
```

## Use this when
- Tape jobs report completed with warnings and the log shows media, read, write, or servo errors.
- Errors appear on some cartridges or in one drive but the library keeps using both.
- You must decide whether the last backup on a warned cartridge can be trusted for restore.

## Not for this skill when
- The tape will not load, eject, or move at all. That is a library mechanics problem.
- Errors are clearly network or source disk errors with no media entries in the log.

## Steps
1. Export the full job log and extract every media related entry with cartridge barcode, drive number, error type, and whether it happened on write or on a later read. Verify: errors are grouped by cartridge and by drive, not left as one warning count
2. Check drive cleaning status, drive error counters, and cartridge health data such as TapeAlert flags, mounts, and bytes written over its life. Verify: cleaning overdue, worn media, and hard error flags are each marked present or absent
3. Clean the drive with a valid cleaning cartridge if any soft error or cleaning flag is present, then run a short write and read test with a scratch cartridge. Verify: the test completes with zero media errors after cleaning
4. Verify the warned backup by reading it back: run the product verify job, or restore a sample of files from the end of the cartridge as well as the start, ideally in a different drive. Verify: sampled files restore and hash or open correctly, or the cartridge is marked untrusted
5. Swap the variables: load a known good cartridge in the suspect drive and the suspect cartridge in a known good drive, and run the same short test. Verify: the error follows the cartridge, follows the drive, or disappears, and that conclusion is recorded
6. Retire or quarantine media that errors in two drives, remove a drive that errors with several good cartridges from the pool, and duplicate any backup that only exists on untrusted media to fresh tape. Verify: the backup catalogue no longer lists the bad cartridge as the sole copy of any required restore point

## Why it happens
Tape errors are often soft at first: a marginal head, dust, or a worn section of tape causes retries that eventually succeed, so the job completes and logs warnings instead of failing. The data written during heavy retry may be readable on the same drive and marginal on any other, which matters years later during a real restore. Error clustering tells the story - one cartridge erroring in every drive is worn or damaged media; every cartridge erroring in one drive is a dirty or failing head; errors only at high throughput can point to the host or interface dropping data. Because tape is usually the last, offline copy, a completed with warnings job quietly lowers confidence in exactly the copy kept for the worst day, which is why read back verification, not job status, decides trust.

## Variants and edge cases
### warnings only near the end of the cartridge
End of tape errors suggest the cartridge is nearly worn out, the drive struggles at the end of the pack, or the job now writes more than the native capacity and compression variation pushes it over. Check native versus written bytes, shorten the job or move to higher capacity media, and retire cartridges that error at the same position repeatedly.
### media errors after a library move or cleaning gap
Dust, a disturbed drive, and long overdue cleaning commonly follow physical moves. Clean all drives, run the library calibration if the vendor provides one, and retest before retiring media in bulk.
### verify passes on same drive, fails on another
That is marginal media or a drive alignment issue, and the backup should be treated as untrusted for disaster recovery where a different drive is likely. Duplicate it to new media now while the original drive can still read it.
Edge cases: some products bury TapeAlert and retry counts in a media or device log separate from the job log, so a clean looking job report can coexist with serious drive warnings - always pull both. Barcode misreads can also attribute errors to the wrong cartridge; cross check the internal media identifier in the catalogue. Keep a per cartridge error history across months: a single soft error after cleaning is survivable, a rising trend is the retirement signal, and waiting for a hard failure means waiting until the data is already hard to recover.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_3mdCyYDcOsb8D8Ynih2EoA
