Hi

 

 

Just as I thought I had come through the teething problems (see other posts) of my system something else pops up to confuse me. Some data points :

 

- I have been systematically replacing my array drives with new WD1002FAEX drives which are SATA3 capable

- These drives I have now found out are a mess of inconsistency ie jumpers bringing them to SATA1 instead of SATA2, labels marked incorrectly etc etc so half the time Im not sure if they have negotiated to sata2 (which is what my cards are)

- I got the config to a point that hdparm -tT /dev/sd* generates read speeds of > 120MB/s in all of the new drives

- I have replaced and am rebuilding the last drive and it is running at ~20MB/s. This looks to be too slow. Note this same drive took 11 hours to preclear which was pretty good based on the benchmarks

 

My config

- Supermicro xSIL-F

- 1 XEON 2.53 quad core

- 4GB RAm

- 3 x AOC-SALP-MV8 cards

- 5 x WD1002FAEX drives (4 Data 1 Parity)

- 1 x WD5000AADS Parity drive

 

I have included my syslog and dmesg. Note the references to UDMA/133?

 

Hope you guys can see something glaringly obvious. I havnt ruled out that its some dumb jumper setting issue.

 

On a separate note if I stop a disk rebuild can I reinsert my old drive? second part of this question can I restart the current drive and will it continue where it left off?

 

Thanks in advance.

syslog_dmesg.zip

An update :

 

- I am now convinced my problem is related to the drives. So for anyone out there although these drives are fast, great warrantee, good price etc I would suggest you steer clear of them to save yourself some frustration (you may never have any issues eg maybe your not using the sata cards I am etc but from what I have seen WD has lost the plot on these ones and they are not worth the trouble)

- As per above comment some are labeled incorrectly, some have a firmware issue where you need to run a specific program over them to correct it, there are two batches, some go to 1.5G interface when you jumper the documented pins 5 and 6, some go to 3G etc etc

- I spent a bit of time last night cancelling the rebuild of the disk and trying different combination of jumper etc and finally got to a combination that generated an ~80MB/s drive rebuild at the start and an average of ~55MB/s overall. I cant complain about this.

 

Now a couple of followup questions:

Q1 When the disk was rebuilding (and it has done this on all of the disk rebuilds, but not on any of the normal parity checks I have performed over the last couple of weeks) there were some writes to all the disks and not just the disk being rebuilt eg parity had 51 writes, and the other disks respectively 15, 11, 4331743 (the drive being rebuilt), 15 and the cache disk 11. There were no errors and the small amount of writes occurred up front in the rebuild. I was under the impression that during a rebuild only the target drive was written to?

Q2  As I was rebuilding this disk and trying to correct the performance issue, I would start the rebuild, let it run for a little while, check the speed, cancel the rebuild, try some tweaks, and restart the machine. The rebuild recommenced automatically (note I didnt re-preclear/zero the disk). I would have done this a handful of times. Is this a problem? Do I have anything to worry about from a corruption point of view or does the rebuild simply wipe the target disk whether every block is zeroed or not?

 

Thanks

 

Q1: Not sure. I think those few writes are normal and should not effect data files.

 

Q2: The rebuild starts over each time. The entire disk is written. This is no problem.