Testers needed for SCSI/SAS spinup/spindown


Recommended Posts

If you have:

1) IBM Br10i - PCI-E x8 Controller: LSI SAS1068e, supports 3 Gb/s SAS/SATA - eight-port

2) LSI SAS 9211-8i - PCI-E x8 Controller: LSI SAS2008, supports SAS 6 Gb/s SAS/SATA - eight-port

or equivalent, operating in HBA mode, and your hard disk spinup/spindown isn't working correctly, please drop me an email and I'll give you a link to a build that tries to fix this.

[email protected]

Link to comment

I have a SuperMicro SASLP-MV8 (PCI-E x4 SAS controller) that I recently added to my system.  Since adding it, I have noticed that only two of my seven hard drives will spin down on their own, no matter how long they've been unused.  Note that six of the seven drives are connected to other SATA PCI-E cards or the motherboard, and all of them were working properly prior to adding the SAS card.  I've been pulling my hair out trying to figure out why they won't spin down properly anymore.

 

If you think this might be related, I'll send an email.    I'm using 5.0b6a, but it is a production server, so I may not be able to help test if you think it's risky.

Link to comment

I have a SuperMicro SASLP-MV8 (PCI-E x4 SAS controller) that I recently added to my system.  Since adding it, I have noticed that only two of my seven hard drives will spin down on their own, no matter how long they've been unused.  Note that six of the seven drives are connected to other SATA PCI-E cards or the motherboard, and all of them were working properly prior to adding the SAS card.  I've been pulling my hair out trying to figure out why they won't spin down properly anymore.

 

If you think this might be related, I'll send an email.    I'm using 5.0b6a, but it is a production server, so I may not be able to help test if you think it's risky.

 

Giz,

 

Have you seen this?

http://lime-technology.com/forum/index.php?topic=12373.0

Link to comment

No, I hadn't seen that.  Thanks for the suggestion.

 

I just changed my spindown times on all my disks from "Use default" to a 3 HR setting (which is what the default actually is), and 3 drives spun down immediately.  The drives that are still up are likely to have been used recently, as I've been working on the server.  Looks like the bug is related to this "Use default" setting, and not the SAS card Tom is investigating.

 

It was so similar I thought it was related, sorry for polluting the thread.

Link to comment

I'm using a Supermicro USAS2-L8i card (8 port 6Gb/s SAS/SATA, using the LSI 2008 chip), running unRAID v5.0beta6.  Although the card is flashed with IR firmware, I believe that it's operating in JBOD mode:

May 23 01:30:06 Tower kernel: mpt2sas0: LSISAS2008: FWVersion(07.00.00.00), ChipRevision(0x02), BiosVersion(07.11.00.00) (Drive related)
May 23 01:30:06 Tower kernel: mpt2sas0: Protocol=(Initiator,Target), Capabilities=(Raid,TLR,EEDP,Snapshot Buffer,Diag Trace Buffer,Task Set Full,NCQ) (Drive related)

 

unRAID believes that it's spinning the drives down, and the green balls flash.  However, every spindown on a drive attached to the Supermicro card produces an error in the syslog:

May 23 01:52:13 Tower emhttp: mdcmd: write: Invalid argument (Other emhttp)
May 23 01:52:13 Tower emhttp: mdcmd: write: Invalid argument (Other emhttp)
May 23 01:52:13 Tower kernel: mdcmd (76): spindown 1 (Routine)
May 23 01:52:13 Tower kernel: md: disk1: ATA_OP_STANDBYNOW1 ioctl error: -22 (Errors)
May 23 01:52:13 Tower kernel: mdcmd (77): spindown 2 (Routine)
May 23 01:52:13 Tower kernel: md: disk2: ATA_OP_STANDBYNOW1 ioctl error: -22 (Errors)

 

Is this related to the problem that you're trying to fix?

Link to comment

Spindown will be released trough internal hdd sleep timer (hdparm  -Sxxx)? If yes, do not forget to re-program timer after return from S3-sleep mode :)

Some words to WD CAVIAR GREEN 1(?2)TB Drives: internal spindown timer blockd on this hdd by permanent execution "smart offline collection" test.

Workaround: execute "smartctl -o off /dev/hdX for all WD Green once, if no spindown, and power cycle system.

 

Link to comment

Spindown will be released trough internal hdd sleep timer (hdparm  -Sxxx)? If yes, do not forget to re-program timer after return from S3-sleep mode :)

Some words to WD CAVIAR GREEN 1(?2)TB Drives: internal spindown timer blockd on this hdd by permanent execution "smart offline collection" test.

Workaround: execute "smartctl -o off /dev/hdX for all WD Green once, if no spindown, and power cycle system.

 

 

Spin-down in unRaid does not use the internal timer feature.  This is because in order to implement Spinup Groups, the unraid driver must keep track of spinning status and be able to check for and spinup drives directly in the driver I/O path.  That is, say disk1,disk2,disk3 are in same spinup group, all spun down.  When driver "sees" read or write to any of those disks, it must also spinup the other disks in the group.  To poll disk spinning status on each I/O would kill I/O performance, hence driver just checks a flag for each disk.  This flag is maintained by having the driver issue spinups and spindowns.

 

Historically, unRaid has also not supported SCSI, mainly because there are a gazillion scsi controllers out there, each with various quirks, and besides, SCSI hard drives have historically been far more expensive (for whatever reason) than IDE and SATA drives, and thus not a big factor in unRaid target market.

 

But with SATA the situation has changed a bit, mainly because of the "libata" layer in linux invented mainly to support SATA drives using the SCSI I/O subsystem.  Since many SAS (Seraially-Attached SCSI) controllers also support SATA (Serial-ATA) drives (in addition to supporting true-SAS drives, yes they exist), we are seeing more SAS controllers in unRaid configurations.

 

Anyway, to make a long boring story shorter, the unRaid driver has used ATA commands to spinup/spindown drives; which are supported by some SAS controllers but not all.  So I added code similar to what's found in hdparm command to also use "SG_IO" interface to send down spinup/spindown commands.

 

This change has been verified with various LSI controllers, thanks to the testers who volunteered, and will be included in upcoming -beta7.  Thanks to all you guys!!

Link to comment

Any way to get this test release - I was away and then got swamped at work.

 

Have a couple BR10i, one Dell 6i/R flashed with older LSI firmware and another still with the Dell's (these are listed as supporting 3TB), and then have one M1015 (SAS2008 based).

Link to comment

Any way to get this test release - I was away and then got swamped at work.

 

Have a couple BR10i, one Dell 6i/R flashed with older LSI firmware and another still with the Dell's (these are listed as supporting 3TB), and then have one M1015 (SAS2008 based).

 

Contact Tom (contact info provide by him at the very top of this post) see if he is willing at this point.

Link to comment
  • 2 months later...

Hi, My 2 br10i's spin down when i select the button to do it manually, but dont seem to spin down after the timer has expired (currently set to 15mins)

 

I have content on the array but I do not use it at present (waiting for shiny new licensed usb sticks), a little nervous now has drives over heated today (controllers screaming at me  :D) even though not being used.

 

Any ideas?

 

Running beta 10.

Link to comment

Hi, My 2 br10i's spin down when i select the button to do it manually, but dont seem to spin down after the timer has expired (currently set to 15mins)

 

I have content on the array but I do not use it at present (waiting for shiny new licensed usb sticks), a little nervous now has drives over heated today (controllers screaming at me  :D) even though not being used.

 

Any ideas?

 

Running beta 10.

 

Your post details about 5.0Beta10 in the wrong post. All spinup/down support on SAS controllers has been fixed and should be working just fine. You should really look into why your drives are overheating though. Ex. not enought fans/placement, location of the server, etc...

Link to comment
Guest
This topic is now closed to further replies.