Ok,

 

Time to add new larger disks.

 

Prior to these new disks, I have various 1TB, 2TB and 3 TBs in my array.  and my parity of course is 3TB.

 

Now, i got two new 6TB WD drives, that i want to add. Since the parity has to be replaced also  (has to be at least the size of the largest disk in the array), what is the exact procedure to replace the parity drive?

 

Moreover, I want to use the second 6TB to replace an exiting 1TB drive.

 

So how can i achieve it, properly as per the procedure?

 

thanks

Do it in two stages ...

 

But before you start, do a parity check to confirm everything is okay before you begin [No disk errors and no sync errors].    If the parity check results in any sync corrections, do it again and confirm that you have all zeroes (both disk errors and sync errors) before you start.

 

Then do the following:

 

(1)  Replace the parity drive and wait for the new sync to finish.    When it's done, run yet another parity check to confirm all went well.

 

(2)  Now replace your 1TB drive with the other 6TB drive and let the system rebuild it.

 

Done 

... although I'd also do a parity check after the rebuild finishes just to yet-again confirm that all went well.

 

Now, i got two new 6TB WD drives, that i want to add.

Have they both passed at least one preclear cycle without errors?

As jonathanm noted, it IS a good idea to test the drives before you use them.  The pre-clear script is a good way to do that.

 

Do you want to keep the parity drive but use as a data drive?

 

Refer to the "Replace parity" link in my sig (bottom right) for instructions to do the type of upgrade you are trying to do.

Now, i got two new 6TB WD drives, that i want to add.

Have they both passed at least one preclear cycle without errors?

 

I haven't unpacked the new drives yet, so no, haven't done any preclears.

 

I have a restriction as of this moment in my array, the card supports a maximum of 16 drives attached, and its full with 16 drives at the moment.  So i think the preclear is not an option with my scenario.

 

no?

Just connect the drives to another system and run the manufacturer's diags ... if you run the Quick test; then do a full Write Zeroes; and then run an Extended test using WD's Data Lifeguard that's a very good confirmation that the drive is good.

 

If possible, do it with a direct SATA connection, although a USB bridge device will work (it will just be MUCH slower).

 

Alternatively, if you have a spare system, you could just connect the drives to it; boot with a basic UnRAID flash key with the pre-clear script on it; and use that system to do the pre-clears.    But since a pre-clear isn't going to save any time (since you're not adding the drives to the array, but are rebuilding existing drives), just using the Data Lifeguard tests accomplishes effectively the same thing as long as you look at the SMART data both before and after the tests and see if anything changes.

 

 

Alternatively, if you have a spare system, you could just connect the drives to it; boot with a basic UnRAID flash key with the pre-clear script on it; and use that system to do the pre-clears.    But since a pre-clear isn't going to save any time (since you're not adding the drives to the array, but are rebuilding existing drives), just using the Data Lifeguard tests accomplishes effectively the same thing as long as you look at the SMART data both before and after the tests and see if anything changes.

 

unRAID doesn't do the preclear once you first insert a new HDD by default?

btw,

 

I don't see any errors on any of my disks via the management GUI of unRAID, and my last parity check of the existing parity drive is:

 

Last checked on Tue Oct 21 19:02:49 2014 GST, finding 0 errors.

 

also, parity is showing as valid at this moment.

 

 

I need the fastest, yet the safest method to add the drives.

If you were adding the drives to your array, then yes, UnRAID would clear the drive before adding it.

 

But you're not adding it, so there's no need to clear it -- it simply needs to be rebuild (with parity in the first case;  with the data from your old drive in the 2nd).

 

The FASTEST way to do this is to simply add the drives (one at a time, as I noted earlier).  There's no absolute requirement to test a drive before using it -- it's very likely it's fine anyway ... but testing them can help find infant mortality issues.

 

If you don't want to spend the time for testing them (which does indeed take a LONG time), you can just do the replacements (one at a time).    But if you do this, be SURE that you do a parity check after each drive is rebuilt ... this is essentially a test that everything went well and will confirm the writes for the rebuilds were all good.    I'd also look at the SMART report for each drive after all is done to confirm there are no parameters that seem out-of-line and that you don't have any reallocated sectors.

 

unRAID doesn't do the preclear once you first insert a new HDD by default?

Yes it does - but the array is not usable while unRAID is doing it (which is a long time with a 6TB drive). 

 

If you use the pre-clear script the array downtime is just seconds.  Also the pre-clear script acts as a good stress test on new disks to avoid disks that fail due to early-life-mortality.

unRAID doesn't do the preclear once you first insert a new HDD by default?

Yes it does - but the array is not usable while unRAID is doing it (which is a long time with a 6TB drive). 

 

If you use the pre-clear script the array downtime is just seconds.  Also the pre-clear script acts as a good stress test on new disks to avoid disks that fail due to early-life-mortality.

 

Actually, UnRAID does NOT do a "preclear" => it simply clears the drive.    The whole idea of "preclear" is to clear the drive BEFORE you add it to UnRAID, so the clear operation isn't required.    As noted, the array isn't usable while a drive is clearing, so pre-clearing it can save significant time when adding a drive.

 

But in this case, it's irrelevant, since the operations you're doing wouldn't benefit from a pre-cleared disk.  Running a pre-clear would simply be a way to test the disk before using it ... which you can do other ways, or you can simply skip testing and use the drives -- as long as the rebuilds work correctly and you get clean parity checks after them, you will have effectively tested them anyway [a rebuild writes to the entire disk; and a parity check reads from it ... and if it's good you know everything is okay].

 

 

i did as suggest, the fastest way. I removed old parity and installed the new one, now it's running the parity sync, which is estimated to finish in about 5-6 days (10 MB/s speed, is this slow?).

 

At least, i'm able to access the data on the array while it's syncing.

 

The process is time consuming, this by itself is taking about 6 days, and then again need to run a check which might also take another 5-6 days, and then replace one of my data drives in the array with my second new 6 TB....so imagine how the whole thing will take

i did as suggest, the fastest way. I removed old parity and installed the new one, now it's running the parity sync, which is estimated to finish in about 5-6 days (10 MB/s speed, is this slow?).

This seems very slow - mine averages something like 100MBs (starting faster than that and slowing down as it progresses nearer the centre of the drives) which is about 10 times your speed.

 

At least, i'm able to access the data on the array while it's syncing.

 

The process is time consuming, this by itself is taking about 6 days, and then again need to run a check which might also take another 5-6 days, and then replace one of my data drives in the array with my second new 6 TB....so imagine how the whole thing will take

It might be worth providing a syslog from the currently running sync to see if we can see any obvious issues.

... I removed old parity and installed the new one, now it's running the parity sync, which is estimated to finish in about 5-6 days (10 MB/s speed, is this slow?).

 

Yes, that's VERY slow ... it should be ~ 10 times that fast.    What are your hardware specs for this system?    Are the drives attached to an old PCI disk controller by any chance?    Do you have some older (very small) drives in the system?

 

As long as the 6TB drive is attached to a motherboard SATA port, things should speed up considerably when the parity sync gets past the oldest, slowest drives [i.e. if you have a 500GB drive, then when it passes the 500GB point; if you have a 1TB drive, then when it passes the 1TB point, etc.]

 

A 6TB drive DOES take a long time to sync and to do a parity check ... but it still shouldn't be more than a day.

 

i did as suggest, the fastest way. I removed old parity and installed the new one, now it's running the parity sync, which is estimated to finish in about 5-6 days (10 MB/s speed, is this slow?).

 

At least, i'm able to access the data on the array while it's syncing.

 

The process is time consuming, this by itself is taking about 6 days, and then again need to run a check which might also take another 5-6 days, and then replace one of my data drives in the array with my second new 6 TB....so imagine how the whole thing will take

This is indeed very much slower than it should be. Post a syslog.

really? then i must have an issue.

 

My drives are all connected to a Dell PERC h200 card, with an intel extender.

 

here is my syslog

 

http://www.filehosting.org/file/details/463073/syslog.txt

 

 

anyone? i uploaded the syslog

... I removed old parity and installed the new one, now it's running the parity sync, which is estimated to finish in about 5-6 days (10 MB/s speed, is this slow?).

 

Yes, that's VERY slow ... it should be ~ 10 times that fast.    What are your hardware specs for this system?    Are the drives attached to an old PCI disk controller by any chance?    Do you have some older (very small) drives in the system?

 

As long as the 6TB drive is attached to a motherboard SATA port, things should speed up considerably when the parity sync gets past the oldest, slowest drives [i.e. if you have a 500GB drive, then when it passes the 500GB point; if you have a 1TB drive, then when it passes the 1TB point, etc.]

 

A 6TB drive DOES take a long time to sync and to do a parity check ... but it still shouldn't be more than a day.

 

Here is the specs, http://lime-technology.com/forum/index.php?topic=21470.0, excpt for the RAID card, I later updated it with Dell PERC H200, which supports a maximum of 16 drives, and total drives connected now (including the parity) is 16.

 

It's worth mentioning, that ealier I connected 17 drives, and faced a similar slowness, but later i found that it was my card H200 was supporting a max of 16 drives. so i removed the 17th drive, and the speed increased. I'm in the process of getting a Dell PERC H310, which can support up to 32 drives.

 

If you check the syslogs, you will see a lot of this:

 

Nov 11 10:33:08 Tower kernel: scsi target0:0:13: handle(0x0017), sas_address(0x5001e674641b1ff3), phy(19)
Nov 11 10:33:08 Tower kernel: scsi target0:0:13: enclosure_logical_id(0x5001e674641b1fff), slot(19)
Nov 11 10:33:08 Tower kernel: sd 0:0:13:0: task abort: SUCCESS scmd(f6de73c0)
Nov 11 10:33:44 Tower kernel: sd 0:0:9:0: attempting task abort! scmd(f75c0780)

 

... My drives are all connected to a Dell PERC h200 card, with an intel extender.

 

More details please.  What motherboard; what slot is the h200 in; and exactly what Intel extender are you using?    If you're using a port expander to connect several drives to a single SATA port, that is likely what's slowing things down here.  But without more details, it's difficult to say.

 

i provided the detail above

... My drives are all connected to a Dell PERC h200 card, with an intel extender.

 

More details please.  What motherboard; what slot is the h200 in; and exactly what Intel extender are you using?    If you're using a port expander to connect several drives to a single SATA port, that is likely what's slowing things down here.  But without more details, it's difficult to say.

 

ok, yes, using intel extender, one port from the perc h200 is going to the extender, and the other is directly connected to a single row in the disk array which is connecting 4 drives. the extender is connected via 5 cables to each row of the disks, where each hosts 4 drives.

 

with this setup, is the speed normal?

In what PCIe is the Perc200? X4 or x8 or x1 perhaps? Perferably an x8. Why don't you use the M1015? It seems you have a severe bandwidth problem.