unRAID Server Release 5.0-beta5a Available

[limetech]

Tom, if you wanted to post up some IOCTL spinup/spindown code program trying different parameters or flags for an testing on the LSI1068E/BR10i controllers, I'm willing to try it out and report back the results.

 

Is the ioctl driver perhaps needed in this situation? Would it help to include/load as a module?

 

http://cateee.net/lkddb/web-lkddb/FUSION_CTL.html

 

I noticed it's not being used.

 

from the .config on 5.0b5a:

# CONFIG_FUSION_CTL is not set

 

 

Issue has to do with spinup groups.  To review reason for spinup groups:

Suppose you have two IDE disks on same cable: one is Master one is Slave.  Suppose both are spun down.  Suppose you now start playing movie from say the Master.  Disk will spin up and movie will start.  Now suppose you access the Slave for any reason, perhaps to browse the directory, or maybe to access some other media file.  When I/O command is sent to Slave, entire controller will "hang" waiting for Slave to spin up and complete the command - that is, no command can go to Master while this is happening, hence you see big nasty pause in the movie.

 

To solve this, all spinup/spindown of array disks is handled by unraid-driver via IOCTL calls to the disks. This is because driver must know each disk spinning state, so that when an I/O comes for a disk that is spun-down, driver can issue spin-up to all drives in that drive's spinup group first, then issue I/O to target drive.  This way, any access to a spun down drive causes all spun down drives in that same group to get spun up.  Effect in scenario above is that at same time Master is accessed, driver also spins up Slave, so that subsequent access to Slave will not cause Master to hang.  Got it?

 

So problem is the IOCTL calls, which are different depending on the device type, that is, ATA/SATA use one form of command, SCSI another, and SAS could use either, except (apparently) that controller which insists on using the SCSI version (which the unraid driver doesn't use).  Trouble is, it's not easy for driver to determine the device type!

 

Anyways, I can think of a few ways to solve this, but I need to think it out further...

 

 

[dyrewolfe]

Tom, thanks for the explanation. Makes better sense now. So I suppose a workaround for now is to disable spindown for the drives attached to that controller and (optionally) spindown manually with hdparm.

[larson]

So I made a fresh install of 5b5a, tinkered a bit and then copied my PRO.KEY to the flash (by SMB). The web server died on me. Below is the last bunch of log messages, and attached is the full syslog.  I messed with the workgroup name, spindown time and the system time and timezone. I noticed the crash after I copied the key file.

 

Did you Stop array, copy key file, then re-boot array?  Or just copy key file?  If the latter, try the former and post back if it works.

 

Sorry, but can't quite recall now. I think I was just checking around and planning to reboot when the GUI hung The array wasn't even built (as I said, fresh install, why would I bother assigning disks to the small std array). Maybe copied some files (preclear, unMENU) to the flash before reboot. But I have done reboots since, and haven't had any problems after that time. (Apart from the ones I'm discussing with Joe, but those are something completely different).

 

/Lars Olof

[limetech]

Interesting new issue that cropped up. Tom/others, if you can confirm?

 

I've wiped my config directory and tried from a fresh configuration, still the same issue:

 

1) From the main unRAID screen, click on any of the disks to view the disk settings

2) Try turning on or off AFP or SMB export

3) Click Apply

4) It should revert back to whatever it was set to before.

 

I can't turn off or on any AFP/SMB disk export, it simply doesn't "apply" the changes.

 

I believe 5.0-beta4, this was fine.

 

 

This bug is fixed in 5.0-beta5b.

[limetech]

... unless you like playing with fire in a diked hydrazine storage area.

 

Never LOL with a full mouth of coffee 

[limetech]

Got to work this morning and put beta-5a on a VM (clean, right from zip) and was unable to modify the disk export status for either AFP or SMB.

 

 

Right, reported earlier & fixed in upcoming -beta5b.

[JarDo]

I'm still receiving segfaults when emhttp loads and have no access to the web gui.  My syslog is attached.

 

This is the offending line in the syslog:

Feb 27 21:48:16 UNRAID emhttp: HOME not found

 

It's crashing because there is no internal variable named 'HOME'.  Are you running a completely stock version of unRAID (ie, no add-ons).  If so, please copy the contents of your Flash device to your PC, zip and email as attachment to [email protected].  There is no personal info in any of those files, except for the names of your user shares, users, and passwords (though the passwords are encrypted - if you prefer you can omit files config/passwd and config/smbpasswd).

 

Tom...  That's the clue I needed.  I had a Workgroup configured in Unraid as "@HOME".  I think that special character was what was killing me.  I changed my Workgroup name and now I'm running Beta v5b5b just fine (so far).  Thank you.