AVS-10/4 Now Available to Order

[Pauven]

Did you try adjusting the "tunables" on the 'Settings/Disk Settings' page?

...

Another tunable that might affect this is "Force NCQ disabled". 

 

Tom,

 

I completed my preliminary testing, and so far the results are very promising!  Thank you.

 

Details are here:  http://lime-technology.com/forum/index.php?topic=27460.msg244639#msg244639

 

It would be very nice if there was an "AutoTune" button in unRAID, it could run a script to perform the tests that I just did manually.  A nice to have for the future, and it might even cut down on support requests...

 

I'm performing a full Parity Check now with the best settings from my preliminary results, and after that a full Rebuild to upgrade a drive.

 

-Paul

[garycase]

+1 for an "Autotune" !!!

 

Now I've got to run a bunch of parity checks on my systems to see if either of them can be "supercharged" for their parity checks !!

 

[coupas]

Can you guys summarize and give step by step play on what just took place? You solved the issue with sas2lp? Please cliff note with clear instructions and I will be in your debt.

[jbartlett]

I saw really weird results when I ran the md_sync tune test on my rig with NCQ left off/unchanged. Ran it to 0.5% with 11 disks. I kept getting CPU stalls at 768 which really hurt the speed for all but one test. Oddly enough, 1280 also received CPU stalls in the syslog but did dramatically better than all the other settings by nearly half.

 

Scores got progressively better and 896/124 reported the same.

 

Tip: http://www.grun1.com/utils/timeDiff.cfm is good for computing time differences reported in the syslog.

 

Still need to research what's causing the CPU stalls.....

 

ETA: Not using the "troublesome" cards

[garycase]

Can you guys summarize and give step by step play on what just took place? You solved the issue with sas2lp? Please cliff note with clear instructions and I will be in your debt.

 

Actually the issue Pauven solved is with the RocketRAID 2760a, but it uses the same chip as the SAS2LP (actually it has 4 of them), so the save technique should work with the SAS2LP, although the optimal setting may be different.

 

You can safely just leave NCQ at its default setting (Off) if you want (or you can try it both ways for each different setting);  but the key is adjusting the md_sync_window setting => this is on the Settings tab under Disk Settings.    The default will vary by controller, but the idea is to try different values until you find an optimal setting for your controller.    To test it, run a parity check and time it.  You don't have to run the whole thing ... just run it to a "known" point and then repeat that for each subsequent change to the setting.    The one that gives you the best time is what you want to use.    Adjust the setting in increments of 128.

 

I'd suggest you read the post where Pauven summarizes his final settings:  http://lime-technology.com/forum/index.php?topic=27460.msg244639#msg244639

 

Edit:  Corrected the parameter to modify -- I had inadvertently typed the wrong one !!  Be sure you change the md_sync_window setting !    Thanks to Pauven for bringing this to my attention.

[garycase]

Any update on the ETA of RC13?

Just finished testing everything I wanted to get into -rc13, so it will be tomorrow (or rather later today - Monday).

 

"Later today" is rapidly fading away 

... did you encounter another issue?

 

[mrow]

Any update on the ETA of RC13?

Just finished testing everything I wanted to get into -rc13, so it will be tomorrow (or rather later today - Monday).

 

"Later today" is rapidly fading away 

... did you encounter another issue?

 

Tom is in Carlsbad, CA if I remember correctly so it's only about 3:45 out there. Still getting close though.

[limetech]

Any update on the ETA of RC13?

Just finished testing everything I wanted to get into -rc13, so it will be tomorrow (or rather later today - Monday).

 

"Later today" is rapidly fading away 

... did you encounter another issue?

 

Yes, but it won't hold up the release... preparing now.

[nars]

Please mention it, if it is something we should know... and don't forget realtek driver

 

Edit: Oops, I see you don't need to "don't forget it" anymore, hope kernel one works fine for all chip variations

[garycase]

I was going to mention the Realtek driver as well -- seems it's the "standard "a"-release item that has caused more -a releases than anything else      Glad it's now in the standard set so it's not something so easily forgot anymore !!

 

[Pauven]

Did you try adjusting the "tunables" on the 'Settings/Disk Settings' page?

...

Another tunable that might affect this is "Force NCQ disabled". 

 

I'm performing a full Parity Check now with the best settings from my preliminary results, and after that a full Rebuild to upgrade a drive.

 

Tom,

 

The md_sync_window parameter has resolved my parity check speed issues, and my disk rebuild speed (which was already good) remained good.

 

Not only did this resolve the issue for the HighPoint 2760A, other SAS2LP owners are reporting success as well.

 

A big thank you!

 

If you care to see additional detail:  http://lime-technology.com/forum/index.php?topic=27460.msg244980#msg244980

 

-Paul

[solomon]

Hello,

 

It would be nice to get some running specs of this server.  I have a custom unraid box but it's a bit slow and draws too many Watt's.  What's the draw on the atom box and what kind of copy numbers and parity check numbers are expected ?

 

Thanks !

 

Solomon