# NEW SERVER STORAGE / NVME / RAID AUDIT — 2026-09-26

Status: VERIFIED CURRENT HARDWARE ASSESSMENT

## Executive conclusion

Both NVMe drives are currently classified as:

`WARNING / endurance exceeded, drive still operational`

They are NOT classified as `replacement required` from the currently observed evidence.

The RAID1 arrays are healthy and should be left unchanged. No repartitioning, RAID recreation, disk replacement, or filesystem rebuild is justified by the current audit.

This server may be used as a temporary migration/production host with enhanced monitoring and conservative write behavior. It should not be treated as a final long-term storage platform without continued observation and a later migration plan to healthier hardware.

## NVMe inventory

Both drives:

- model: Samsung `MZVLB1T0HBLR-00000`;
- family: PM981a 1 TB;
- firmware: `EXF7201Q`;
- PCIe: Gen3 x4.

### nvme0

- Critical Warning: `0x04`;
- reliability bit: set;
- Available Spare: 100%;
- Spare Threshold: 10%;
- Percentage Used: 177%;
- Host Written: 614.31 TB;
- Host Read: 129.50 TB;
- Power-on Hours: 39,130 h;
- Power Cycles: 17;
- Unsafe Shutdowns: 3;
- Media and Data Integrity Errors: 0;
- Error Log Entries: 282;
- Temperature: ~35 C;
- Warning temperature time: 0;
- Critical temperature time: 0;
- PCIe fatal/nonfatal/correctable errors: 0.

### nvme1

- Critical Warning: `0x04`;
- reliability bit: set;
- Available Spare: 100%;
- Spare Threshold: 10%;
- Percentage Used: 175%;
- Host Written: 613.31 TB;
- Host Read: 122.18 TB;
- Power-on Hours: 39,156 h;
- Power Cycles: 17;
- Unsafe Shutdowns: 3;
- Media and Data Integrity Errors: 0;
- Error Log Entries: 282;
- Temperature: ~33 C;
- Warning temperature time: 0;
- Critical temperature time: 0;
- PCIe fatal/nonfatal/correctable errors: 0.

## Interpretation

`Percentage Used > 100%` means the controller's endurance estimate has exceeded its nominal design life estimate. It does not by itself prove current media failure.

However `Critical Warning 0x04` indicates the NVMe subsystem reports degraded reliability. Therefore the drives are not considered normal-wear devices even though they continue to operate correctly.

Current evidence against immediate replacement:

- Available Spare remains 100% on both drives;
- Media/Data Integrity Errors = 0 on both drives;
- no read-only bit;
- no temperature warnings;
- no PCIe AER errors;
- no kernel NVMe timeout/controller reset/I/O error evidence observed;
- RAID is not degraded;
- mirror mismatch count is zero.

Therefore the current classification is WARNING, not replacement-required.

## Error log interpretation

Both drives show `Error Log Entries = 282`.

The latest observed error type is:

`Invalid Field in Command`

with:

- LBA = 0;
- NSID = 0.

This is not evidence of a bad LBA, uncorrectable read, or media-integrity failure.

The real media-integrity counter remains zero on both drives.

The value 282 did not increase during the current SMART/self-test investigation.

## NVMe self-test result

Short Device Self-Test was run separately on both drives.

Both returned:

- Operation Result = `0x7`;
- Segment Number = `0x2`.

The result is consistent with the already-present Critical Warning during the SMART-check stage. No separate evidence was found for failure in later NVM/media-integrity stages.

Therefore this self-test result is not treated as proof of NAND/media failure beyond the existing reliability-degraded warning.

## RAID/mdadm state

### md0

- RAID1 `[2/2] [UU]`;
- Failed Devices: 0;
- degraded=0;
- mismatch_cnt=0.

### md1

- RAID1 `[2/2] [UU]`;
- State: clean;
- Failed Devices: 0;
- degraded=0;
- mismatch_cnt=0.

### md2

- RAID1 `[2/2] [UU]`;
- State: clean;
- Active Devices: 2;
- Working Devices: 2;
- Failed Devices: 0;
- degraded=0;
- mismatch_cnt=0;
- internal bitmap present;
- both members have event counter `9332`.

All six RAID member superblocks report correct checksums.

Both sides of every array report `Array State: AA`.

`md2` completed its initial resync after the 2026-09-25 install:

`md: md2: resync done`

No current kernel evidence was observed for:

- I/O error;
- NVMe timeout;
- controller reset;
- EXT4 filesystem error;
- md degraded.

## Current storage decision

Canonical decision as of 2026-09-26:

1. Keep the existing RAID1 layout unchanged.
2. Do not replace either NVMe based solely on the current audit.
3. Do not repartition, recreate md arrays, reformat, or alter filesystem layout during the temporary-host phase.
4. Classify both NVMe devices as WARNING / endurance exceeded, drive still operational.
5. Add enhanced SMART/NVMe/md monitoring.
6. Keep external backups because both mirrored devices have similar age/endurance history and therefore share a correlated wear profile.
7. Avoid unnecessary heavy write amplification/backfill while this host is temporary.
8. Plan a later final migration to healthier storage/hardware.

## Escalation criteria: WARNING -> REPLACEMENT REQUIRED

Escalate a drive to `replacement required` if one or more serious indicators appear, especially if persistent or increasing:

- `Media and Data Integrity Errors > 0` and increasing;
- Available Spare starts falling materially or crosses the threshold;
- read-only critical-warning bit appears;
- NVMe timeout/controller-reset/I/O errors appear in the kernel;
- PCIe AER errors begin to accumulate;
- any md array becomes degraded;
- RAID mismatch appears after a valid consistency check;
- self-test begins failing on NVM/media-integrity stages rather than only the already-known SMART warning stage;
- filesystem integrity errors appear;
- repeated unexplained device disappearance/reset behavior.

## Consistency-check note

A full ~1 TB md RAID consistency patrol read has NOT yet been run after this audit.

It is non-destructive but creates a complete read workload on both SSDs. Because the arrays are currently clean, mismatch count is zero, there are no media errors, and the server is being treated as a temporary host, this check is optional rather than an immediate prerequisite.

If performed, record:

- start/end time;
- load impact;
- resulting `mismatch_cnt`;
- any new SMART/NVMe/kernel errors.

## Temporary-host operating policy

For the intended temporary migration-host period:

- retain old/source systems as rollback where practical;
- no single-copy critical data;
- external/off-host backup required for important migrated state;
- monitor SMART/NVMe/md state frequently;
- use early disk-space alerts well before critical fullness;
- avoid unnecessary backfill/rewrite-heavy research jobs;
- do not treat RAID1 as a substitute for backup;
- continue with server/application migration only while the above indicators remain stable.