[Solved] - Agonizingly slow parity, write - everything

[Phil C.]

I'm hoping someone can help.

 

I have an unRAID server that I built in 2010, basically my home version of the one sold by Lime Technology at the time.  Supermicro C2SEA motherboard, Icy Dock cages, 7 drives (now) mostly Seagate 1.5 TBs.  I had a Promise SATA300TX4 Controller, but swapped it for a Supermicro AOC-SASLP-MV8.  I have flashed the MB to the latest BIOS, as well as the latest firmware for the SASLP-MV-8.  Can't get the Promise controller updated - Just too hard to do booting from a USB flash drive with no floppy or CD-ROM.

 

Anyway, the server has ALWAYS been slow.  Parity checks less than 10Mb/s that take 5 and 6 days to complete, but I just lived with it because I didn't know any better.  In the meantime I built a new server to handle media service, and relegated the "old" one to being an archive.

 

Yesterday my parity drive started having write errors.  I had a spare drive (WD 2Tb) so I put it in.  Parity-sync started - and its going at <5Mb/s.  It will take over a week for this to complete.  I let it run for most of today (~8 hours) and it got up to 1.5%.  Just for comparison, I fired up a parity check on my new server; it's running at 95Mb/s.  It will be done in no time.  The new server shows nearly constant disk activity on all the disks at once; the "old" server shows constant activity on the parity drive, but only flashes activity on all the others.

 

I have changed all the cables, changed back to the Promise controller, checked all the BIOS settings - I have done everything I know to do.  I even rolled the unRAID software back to 15A.  Nothing helps.

 

I've attached a syslog for the server, about 20 minutes into parity-sync under RC-16c.  Don't know if it will help but maybe.

 

Would appreciate any ideas anyone has.

 

Phil C.

syslog-2013-07-13.txt

[garycase]

... the server has ALWAYS been slow.  Parity checks less than 10Mb/s that take 5 and 6 days to complete, but I just lived with it because I didn't know any better.

 

Should have isolated this long ago ... you most likely have either (a) a defective SATA port on your motherboard;  (b) a defective Promise card;  © a bad drive; and (d) a bad SATA cable.

 

Have you tried replacing the SATA cables?

 

Do the following:

 

Copy a reasonably large (500MB or larger) file from EACH of the individual drives (one-at-a-time) to your PC, and note the data rate you're getting for each of theses transfers.

 

Are you on a Gb network?

 

[Phil C.]

I appreciate the help.

 

I have changed cables.  I do have a gigabit network - although right now, the server is on an airport express (802.11n) because I had to relocate it out of my basement server room to work on it.  Server was on the gigabit network when I had the initial problem.

 

I tried to copy a file from each drive.  Files were from 400 to 600MB each.  All exhibited approximately the same data rate (completion in 2-3 minutes, difference being what I expected based on difference in file size).  So no sore thumb stuck out.  Of course, it could be that the wireless is slow enough to mask I/O differences...is that possible?

 

Please keep it coming....

Thanks,

Phil

[Phil C.]

Well, sh*t.

 

I had three or four drives on hand (old but hardly used).  When the parity drive failed this morning, I replaced it.  That's what started this day-long extravaganza.  Just changed the parity drive again - and now I'm getting 65-70 MB/s.  Son of hibachi.

 

Two bad drives in a row...and I guess that first one had been bad for a very long time - like, since I bought it.

 

I'd scream, but it would wake the five-year-old...

 

Thanks for the help - after doing the transfers off the data drives, I thought "Hmmmm, what if it's the parity drive, the one drive I can't copy files to/from?"

 

Phil C.

[garycase]

Glad it's resolved ... unfortunate you lived with the low data rates for so long ... but at least it's all okay now; and I'm sure you'll attack quicker if you get a problem in the future