Unable to recover the RAID

[Merlin]

Hello Alexey,

we have the fllowing problem:

we have a RAID 5 of 3 drives + 1 spare
original drives had a lot of bad sectors so we could clone all of them.

the first attempt with ZAR using all drives was not successful and after "Quick identifying data" we'v got a message "Unable to continue..."
Then we tried to use only three drives and the result was the same.

here is the log of the first attempt:


Starting RAID5 parity check
Parity detected in 95% of 200 tests
Raw scanner starting on disk 0102 from LBA 0 to LBA 398297089
Raw scanner starting on disk 0103 from LBA 0 to LBA 398297089
Raw scanner starting on disk 0104 from LBA 0 to LBA 398297089
Raw scanner starting on disk 0105 from LBA 0 to LBA 398297089
__________________________________________________ ______

Performance: Ident data query (72482 of 84816) 0s
RAID: fetch 72482 entries, drive 0102
Performance: Ident data query (61255 of 63973) 0s
RAID: fetch 61255 entries, drive 0103
Performance: Ident data query (47043 of 49171) 0s
RAID: fetch 47043 entries, drive 0104
Performance: Ident data query (72411 of 84818) 0s
RAID: fetch 72411 entries, drive 0105
Volume information:3.0
Volume information:3.0
ALERT!: zero-sized MFT attribute
ALERT!: zero-sized MFT attribute
__________________________________________________ __

ALERT!: Excessive MFT attribute size near ap 269; hint -1
Volume information:3.0
Volume information:3.0
Performance: BuildSwitchList 1m 25s
Vote RAID stripe size: 128; Count=2042 ; separation 77,5%
Purged a total of 13942 known false positives (stripe size mismatched)
Remaining 4401 records
Vote Start offset: 0; Count=4346 ; separation 99,9%
RCRD/RSTR bias @976
RAID: Test RAID0
Performance: Ident data query (12282 of 84816) 0s
Parity for disk 0102 at 1
ALERT!: Too low separation (2980 vs. 3187)
ALERT!: Too low separation (2980 vs. 3021)
ALERT!: Too low separation (2980 vs. 3094)
Performance: Ident data query (2715 of 63973) 0s
Parity for disk 0103 at 0
ALERT!: Too low separation (645 vs. 682)
ALERT!: Too low separation (645 vs. 727)
ALERT!: Too low separation (645 vs. 661)
Performance: Ident data query (2124 of 49171) 0s
Parity for disk 0104 at 1
ALERT!: Too low separation (485 vs. 570)
ALERT!: Too low separation (485 vs. 552)
ALERT!: Too low separation (485 vs. 517)
Performance: Ident data query (12365 of 84818) 0s
Parity for disk 0105 at 1
ALERT!: Too low separation (2997 vs. 3118)
ALERT!: Too low separation (2997 vs. 3099)
ALERT!: Too low separation (2997 vs. 3151)
Parity index conflict between 0102 and 0104
Final nonzero RAID layout scores
UNABLE TO CONTINUE:
Unable to reconstruct the RAID
Unable to continue processing, program will now terminate.
Not a bug - data is probably beyond repair, or something is misconfigured.

is there something we should try? what would be your advice?

thank you so much,
Larisa

[Alexey V. Gubin]

what is the build number you are using for this run, as indicated on the first screen?

[Merlin]

the build number is 8.3-36

thanks,
Larisa

[Alexey V. Gubin]

I've been reviewing this part of ZAR for quite a long time now, suspecting that something may be amiss, but still not able to confirm there is a problem, let alone isolating it.

This type of the error (parity index conflict) is triggered when ZAR identifies two drives in the array having parity in the same row (which is not possible on a valid RAID5; only one parity per row is allowed). This may occur if you swap two drives in the array and then do a rebuild on it (assuming controller does not recognize the fact the drives were swapped).

We keep reviewing this part, but there is nothing I can suggest for now; and also no certain timeframe on the fix (supposing there can be a fix).