February 18, 201115 yr What should my next step be? Disregard the parity drive and unassign it from the array, install a new drive, then build parity? Media errors are un-readable sectors. You can un-assign the parity drive Then start the array with it un-assigned Then stop the array Then re-assign it. Then start the array once more. It will completely write parity to the disk, this should re-allocate those un-readable sectors Then perform another parity check. (to read what was just written) There should be no errors. Then get a smart report for the parity disk. (To see its general health and to see how many sectors were re-allocated) If too many sectors were re-allocated for comfort, or more are pending re-allocation, it is time for a new parity disk. Joe L.
February 18, 201115 yr Oy vey! I just cringe at all the days it will take to do this ... Perhaps I missed something? Why days?
February 18, 201115 yr Author It takes roughly 10 or so for a parity check/rebuild. Depending upon when I start the process, I usually don't wake myself up in the middle of the night when it completes to begin another process. So after I'm done moving a backlog of files over (the parity is already suspect and must be rebuilt anyways), I will start the parity rebuild later this afternoon. If all goes well, tomorrow morning I will then begin the parity check, ending in the evening. Two days at least, if everything works out...
February 18, 201115 yr It takes roughly 10 or so for a parity check/rebuild. Depending upon when I start the process, I usually don't wake myself up in the middle of the night when it completes to begin another process. So after I'm done moving a backlog of files over (the parity is already suspect and must be rebuilt anyways), I will start the parity rebuild later this afternoon. If all goes well, tomorrow morning I will then begin the parity check, ending in the evening. Two days at least, if everything works out... OK, now I understand. 10 hours is not unusual. My older array takes much longer. I too sleep through the bulk of it.
February 20, 201115 yr Author UPDATE: Wow, I can't believe it, but FINALLY, NO ERRORS! It makes me wonder, however, if all the weeks exhausted on troubleshooting, the money spent on "redundant" components (new break-out cable, new hard drive, new PSU), breakdown for supposed "repair", if that was all for naught it all it it required was to rebuild the parity drive. Because even after the "repair", after all the replacements, I still kept getting the same errors on the ata7 channel. But simply rebuilding parity everything is now cleared up and zero errors anywhere; even the ata7 channel. Doesn't make sense. The possible clue was the lack of consistent error reporting in regards to the parity drive. Which does leave me a little uneasy with the accuracy of unRAID's error reporting and detection. Now comes the waiting game to see how long this setup will last with no errors...
February 20, 201115 yr UPDATE: Wow, I can't believe it, but FINALLY, NO ERRORS! It makes me wonder, however, if all the weeks exhausted on troubleshooting, the money spent on "redundant" components (new break-out cable, new hard drive, new PSU), breakdown for supposed "repair", if that was all for naught it all it it required was to rebuild the parity drive. Because even after the "repair", after all the replacements, I still kept getting the same errors on the ata7 channel. But simply rebuilding parity everything is now cleared up and zero errors anywhere; even the ata7 channel. Doesn't make sense. The possible clue was the lack of consistent error reporting in regards to the parity drive. Which does leave me a little uneasy with the accuracy of unRAID's error reporting and detection. Now comes the waiting game to see how long this setup will last with no errors... A parity error indicates a single "bit" on one of your disks was not as expected. Unfortunately, a count of all the bits across an address that results in an odd number does not tell you which bit needs to be flipped on which disk to end up with an even number of "1" bits set. It also does not tell you if the bit was stored incorrectly on a given disk or read incorrectly, or the math in memory was corrupted in the MB chipset. You are faced with a process of elimination. I'll illistrate. LEt's assume I pick a single bit from address 100000 on all my 11 of my disks. The values I see are 10000110101 Now I read parity and see the bit there is a 0 Now, as you see, there are 5 bits set to a 1, and this is an error since I should have an even number of bits set to a 1. I could fix it by flipping ANY bit to its opposite value. Problem is, no way to know which is incorrect. unRAID reports the errors just fine. All it can do is report a parity error (an odd number of bits set to a 1) or a read error (a specific disk itself failed to return data when read) or a write error (a specific disk returned an error when written to) The "parity" errors are usually from bits that were not written to the disk, usually when power is turned off without stopping the array first. parity errors can occur from almost anything though, a disk that was written a "1" but reads, a "0". Bad RAM, noisy power supplies, bad disk controllers, bad motherboard chipsets. (The Nforce4 chipsets are known to cause random errors) I'm glad you are not seeing errors now, but it might not have been the parity disk, or perhaps it was... It could be the data was originally written poorly, and read back with difficulty, so re-constructing it might have re=written the data and you'll be fine. I'd suggest regular parity checks... The NOCORRECT type, until you have more confidence in the array.
Archived
This topic is now archived and is closed to further replies.