N3200 pro / 1.00.05c HDD problem...

[RaMoNsTeR]

Hi,

Upgraded to 1.00.05c firmware that was just released from the old 03 series firmware (last available before new one). Here is the problem: HDD's used to spin down and power save just fine previously, but after update no matter what they are always on. Tried toggling config settings to no avail. I have 3 x WD20EADS 2Tb Western Digitals, which have been fine until FW update.

Pretty serious bug there in my opinion, the only thing thats changed is the firmware, even with all other devices turned off on the network the drives will continue to spin endlessly so its not an issue of media library scans or some other network activity, anyway it was fine with the previous firmware.

Thecus need to fix this dire and very serious bug ASAP as this contributes very negatively to overall longevity and reliability of the drives, and this is what a lot of us purchase these units for - the raid 5 piece of mind. If we are talking about going months without an update or fix I might as well return the N3200 pro unit now and go for a QNAP or similar. Have raised a ticket directly with Thecus and will be eagerly watching and waiting to see the response which will govern the ongoing enthusiasm and kinds words or otherwise I have for the products.

@Thecus: Please fix this ASAP and release a bugfix firmware.

[RaMoNsTeR]

EDIT: removed, not solved, see below...

[RaMoNsTeR]

/bump - although it appeared a config reset fixed this issue - it must have been blind luck that allowed the drives to "standby" in that instance, no luck since. I have started investigating this issue heavily and so far have discovered the following with regards to my N3200 pro and 3 x WD20EADS 2Tb Western Digital HDDs

- All services/extras apart from NFS and SMB turned OFF. The ONLY module installed is DROPBEAR to access the SSH.

- Have unplugged NAS from network for extended period and drives were still spinning.

- hdparm -y option will put the drives into standby successfully every time.

- Turning off power management using -B 255 then setting a very fast spindown (option -S) time does not work... so we assume theres a minimum set, generally this is 30 minutes.

- As per ATA standard, setting anything under 0x80 (128) i.e. 1-127 causes the power management to allow not only drive head parking (8 seconds usually) but also spin down. Confirmed, this is working.

- Setting -B from 128 to 253 does park the head quickly and cause an increment of the load cycle in SMART data.

- Setting -B 254 definitely turns off this quick head park behaviour causing large numbers seen in the load cycle, these drives are rated at 300,000 cycles... assuming we can get standby to work after 30 minutes it *might* be enviable for some users to use -B 254 over -B 253 i.e. users that access the drives a lot randomly during the day vs users that access 1-2 off to watch a long movie only every so often.

- 100% confirmed THIS IS THE IMPORTANT BIT THECUS!!! -> Setting the head to park using -B 253 (and therefore cause an increment of the load cycle number in SMART if accessed) and setting -S 241 which is what N3200 pro does for 30 minute standby setting THE DRIVE WILL NEVER STANDBY... something added in the new firmware (smart checks???) causes access around every 30 minutes or so and causes the drive to never reach 30m idle and spindown!!! You can PROVE this by seeing the load cycle SMART data has incremented by 2 when you check it , once for the access and once more for your smart check!!!!! ARGHH!!

- Curiously: if you force the drive to spindown using -y, it will 100% spin down and behave and whatever was causing the drive to access and not allow an idle 30m spindown will in fact NOT wake it up!!! This is common of SMART probes like -C or -A and other such checks, or I believe if the HDD is on standby (spindown) the kernel is smart enough to provide cache or delay whatever is causing the lack of spin down normally. I have some diagnostic tools (from QNAP ftp) and scripts which I will continue to monitor to try and work out what is causing this.

- This is NOT a fault of the drives APM as *something* causes access and increments the load cycle smart counter when the heads are fast parked prior to a (hopeful) 30m spindown, its something on the NAS.

- This problem seems more specific / more severe on the latest 05c firmwares, drives used to park most times on the previous 03.

Thoughts? Comments from Thecus... its definitely a problem.

Also I suggest owners of WD20EADS set the accoustic setting to 128, seems like its set to 'fast' 254 which isn't needed for this situation.

[RaMoNsTeR]

/bump - no time to investigate further as yet, no one else having spindown issues on 05c firmware?

[poolshark]

No problems here, but I have Seagate disks. Have you tried downgrading the firmware ?

[RaMoNsTeR]

ATT: N3200 pro w/Firmware 1.00.05c does not put drives 2 and 3 to standby mode after the specified time! specific to WD20EADS? Unlikely, read on...


Ok I have carried out some testing and concluded there is some low level (kernel/unlogged/controller) activity and/or issue which does not allow the N3200 pro running 1.00.05c firmware to shut down the 2nd and 3rd drives after the configured spin down time (usually 30m), only the drive designated /dev/sda is put to standby!

Please refer to the following output from the unit: http://rapidshare.com/files/32121632...dby_issues.zip

I've included Parse_showing_checker_works.txt to show disk activity is logging fine as illustrated by smbd traffic.

The HDD_standby_issue_N3200pro.txt file shows a 1 hour test monitoring drive access using by setting /proc/sys/vm/block_dump to 1 and finding relevant activity by parsing dmesg -c. You can see the first HDD goes to complete standby at the 28 minute mark, plus we have waited 2 minutes at the start for xfssyncd and pdflush transient activity to pass before logging, so 'sda' spun down exactly on the correct 30 minute mark as specified at the start of the test using hdparm -B 253 -S 241 /dev/sda /dev/sdb /dev/sdc

Alarmingly you can see at the same point and throughout the 1 hour test run which parses and checks for activity and drive state every 30 seconds that my 2nd and 3rd drives DO NOT successfully spin down although drive 1 (/dev/sda) does! Absolutely NO activity is logged, equally strange is the larger Load_Cycle_Count on /dev/sda at the end, although this illustrates *ALL* drives had some form of low level activity during the test cycle that was not shown through the system block dump method.

hdparm -y can successfully standby all 3 drives as does any -B option under 0x80 (1-127) as per the ATA standard. The drives are running the same firmware as shown here:

/dev/sda:
Device Model: WDC WD20EADS-00S2B0
Serial Number: WD-WCAVY0737077
Firmware Version: 04.05G04

/dev/sdb:
Device Model: WDC WD20EADS-00S2B0
Serial Number: WD-WCAVY0750934
Firmware Version: 04.05G04

/dev/sdc: Device Model: WDC WD20EADS-00S2B0
Serial Number: WD-WCAVY0738697
Firmware Version: 04.05G04

Same firmware, in fact the disk that does standby is only serial stamped 1500 or so prior to /dev/sdc, I don't believe its the drives - but the unit on the controller/firmware or kernel level "upsetting" these drives or failing to pass through the correct spin down parameter from /etc/cfg/rc at startup via hdparm. Seemed like the unit would park better on previous firmwares as mentioned.

I can't do much more at this point, PLEASE Thecus come to the party, check your kernel sources and configs etc and solve why 03 firmwares standby drives better than the 05, or at least get the sdb and sdc spinning down!

[RaMoNsTeR]

Got hold of 3 x Seagate 1.5Tb drives to test, raid finished building late last night, and surprise surprise only the first drive spins down after 30 minutes. My ticket has been closed with no response from Thecus - I'm just wondering now how I go about getting a return # for this thing so I can dump it back to the distributor and purchase one of the other working options out there

[zdrux]

I just bought my N3200Pro and also noticed this after upgrading my firmware. I wish I had known this earlier, then I wouldn't have flashed it

Any progress on fixing this?

[RaMoNsTeR]

Yeah I fixed it, ebay'd the Thecus and purchased a QNAP TS-659 - software and performance are in a totally different league, the quality difference is marked - and the drives shut down in a timely fashion. I know theres competition between these two companies in the low/mid sector but until now I thought they were much the same, but the polish of firmware/product and customer service response time and information given by QNAP is significantly better than Thecus.

Sorry to be the bearer of bad news zdrux, but they won't fix it, they shutdown sometimes but its highly unreliable, my drives had a huge disparity in their SMART reported running times. If you just bought it return it and get the comparative QNAP you won't look back.