init: F3 unlock project — ESP32-C3 bridge + procedure for CLAR200 Exos X18
This commit is contained in:
@@ -0,0 +1,41 @@
|
||||
# IDEA.md — Project Idea
|
||||
|
||||
## What is this?
|
||||
|
||||
A procedure + ESP32-C3 firmware to talk to the **Seagate F3 diagnostic serial
|
||||
terminal** on two EMC/CLARiiON-locked Seagate Exos X18 SAS drives
|
||||
(ST16000NM004J, EMC firmware ES2264), in order to reformat them out of the
|
||||
locked "SS160520 CLAR200" state and back into usable 512-byte-sector drives
|
||||
for the `tank` ZFS pool.
|
||||
|
||||
The two drives (SN **ZRS015V7**, **ZR511KWK**) were bought from ServerPartDeals
|
||||
already pulled from an EMC array and never properly decommissioned. They answer
|
||||
SCSI INQUIRY/VPD but return `Hardware Error / Internal target failure (44h/00h)`
|
||||
to every other command — including READ CAPACITY — so **no host-side software
|
||||
(sg_format, openSeaChest, SeaChest, firmware download) can touch them.** That is
|
||||
fully documented in `~/bin/EMC_CLAR200_DRIVE_FIX_NOTES.md`.
|
||||
|
||||
The only remaining DIY lever is the drive's **F3 service terminal**, an
|
||||
out-of-band UART on the PCB that bypasses the SCSI command layer entirely. This
|
||||
project builds the cable (an ESP32-C3 acting as a 3.3 V USB-serial bridge) and
|
||||
documents the F3 procedure to attempt the unlock.
|
||||
|
||||
## Problem it solves
|
||||
|
||||
- `tank` is short on **14.6 TB+** spare disks: mirror-35 has a dead + a dying
|
||||
leg, and ex-mirror-38 is a bare single-disk vdev. Both need 14.6 TB disks.
|
||||
- These two CLAR200 drives are 16 TB and would fix both holes — *if* they can be
|
||||
unlocked. Otherwise they're paperweights / RMA candidates.
|
||||
|
||||
## Constraints / preferences
|
||||
|
||||
- **Cable:** ESP32-C3 already attached to steel141 at `/dev/ttyACM0`
|
||||
(native USB-Serial-JTAG, MAC `58:8c:81:ac:f0:4c`). 3.3 V logic — native match
|
||||
for the F3 header. No separate USB-TTL adapter needed.
|
||||
- **Drive location:** ZRS015V7 currently in a JBOD on pve173. F3 work must be
|
||||
done **on a bench**, not in-chassis — never risk the live 549 TB tank.
|
||||
- **Honesty:** Exos-X18 F3 dialect ≠ the well-documented 7200.11 era. The
|
||||
procedure discovers commands via the drive's own `?` help; it does not blindly
|
||||
fire format commands copied from old forums.
|
||||
- **Fallback if F3 fails:** RMA to ServerPartDeals (sold not-decommissioned),
|
||||
or decommission in a real EMC VNX array, or PC-3000 SAS.
|
||||
Reference in New Issue
Block a user