MoxFalder wrote:
- i use pci sata card "Promise FastTrak 378"
This
might be involved with your reported symptom of "dropped partition" and it is causing problems with diagnosis - see below. If I was in your situation, I would
immediately stop using this HBA, and use a normal (e.g. on-board, non-RAID, standard ACPI or even IDE-emulation) HBA, until you have found the cause(s) of your problem (or, of course, your problems might stop after you use a normal HBA
).
SMART data shows the drive is not perfect, but much better than I had expected in my hypothesis - however I thought you had done a full Windows format recently, in which case there should be zero sectors pending reallocation, but...:
MoxFalder wrote:
Code:
197 Current_Pending_Sector 0x0032 200 200 000 Old_age Always - 1
198 Offline_Uncorrectable 0x0030 200 200 000 Old_age Offline - 2
... that is therefore a surprise, if the drive has just been zero-filled.
This next attribute suggests there have been SATA errors at some point in the past:
MoxFalder wrote:
Code:
199 UDMA_CRC_Error_Count 0x0032 200 200 000 Old_age Always - 13
... but without seeing previous SMART data for comparison, we don't know
when this value was increasing, and
when (or if!) it has stopped increasing. I see you mention that you have replaced the SATA cable recently. However SATA cables are not the only cause for this counter to increase. Further monitoring of this counter is needed IMHO.
MoxFalder wrote:
Code:
Warning: ATA error count 11617 inconsistent with error log pointer 1
This is bad news in 2 ways - 11617 logged errors is a
big number, but due to an apparent problem with reading the drive's internal log, we cannot see what commands were failing and exactly how they were failing. This is very bad news for diagnosis.
I
guess that you still had the WD drive attached to that Promise HBA, when you ran smartctl to gather the SMART data you supplied. If so, then smartctl probably used a pass-through command to try to gather the SMART data from the drive, even though the HBA stopped WD DLG from getting this direct access (see below). Again, I would move that drive to a "dumb" HBA and re-run the gathering of SMART data, to see if the problem reading the SMART ATA error log is caused by the pass-through functionality needed when attached to the Promise HBA, or is actually a drive problem.
FYI this SMART data was gathered with no
recent SMART tests having been done on the drive (last SMART test was done 1623 power-on hours before this SMART data was collected on 21 Feb), and that was not a full test, but a "conveyance" test.
From dlgdiag:
Code:
Test Option: EXTENDED TEST
Model Number: Promise 1+0 Stripe/RAID0 [<<< This shows test was NOT accessing the drive directly]
Unit Serial Number: [<<< Serial number missing for same reason]
Firmware Number: 1.10
Capacity: 1500.30 GB
SMART Status: Not Available [<<< SMART status not available for the same reason, which prevented the main reason for running DLG, which is that it can perform proprietary checks of WD drives, to give greater confidence of correct diagnosis]
This is typical when a RAID HBA is used, and it prevents normal direct access to the individual drives, so that is affecting the testing which DLG can do.
MoxFalder wrote:
Also under gparted i made a "check" partition and got this strange result
I didn't ask for that test, and I don't have time to review all that data at the moment. You can ask whoever suggested that test, to analyse the result for you. My brief suggestion would be to review the Linux dmesg (or /var/messages, or however you can see the Linux messages when gparted was being run) to see if there were any errors being logged. The errors from gparted
seem to suggest logical, not physical, problems and again that could be related to the HBA and/or its drivers (but the Linux kernel messages do also need to be checked).
Therefore yet again, to eliminate any problems being introduced by that HBA or its drivers,
I would use a standard HBA instead for all testing now, until you are confident in the drive itself. Then re-run SMART data collection and re-run DLG tests, after confirming that the drive model is correctly recognised by DLG.
FAQ
Search
Login
Register
HDDGURU FILES
Remember to collect the full, raw, SMART data. Some versions of WD DLG do not show that detail of the SMART data, but smartmontools does (as well as many other utilities). You need to run WD DLG for its testing, even though its SMART data reporting is not the best.