Hello all.

 

Completely new to the Unraid experience. I've recently decided to build a new server and managed to get some nice hardware that a company was no longer using. Everything seems to be going fine, but I am seeing some very slow parity speeds and can't seem to figure out what is going on. I have trawled through these forums and absorbed a massive amount of data, it really helped me get my system off the ground. Been lurking and learning ever since, but I seem to have hit a roadblock.

 

I've got 8 1TB WD Red 2.5" drives in my array, and 2 2TB WD Blue 2.5" drives as parity (the gear I have only supports 2.5" drives via a Dell PERC H730P card running in HBA mode). All drives are less than two weeks old and individually run at good speeds. When the parity job starts, it looks great, over 100MB/s and an estimated process time of 6ish hours.

 

Within 20-30 mins though, the speed slows to a crawl and the estimated time goes from hours to days. I've tried stopping, rebooting, and kicking off the process again with the same results.

 

 

 

I've set the md_num_stripes tunable to 4096 based on a number of posts in the forum, doesn't seem to have made a difference. I've disabled everything I could find including docker, nothing should be taxing these drives at all apart from the parity build. I know the parity drives are not the best model, but could that explain such terrible performance?

 

Turbo write is enabled, all other settings are at defaults. I'm also wondering if it's the PERC card that is making things difficult, all the other hardware should be resulting in pretty good performance, and it's the only bit of hardware that I am not entirely familiar with. Are there any special settings within Unraid or even the BIOS (Dell Poweredge T630) that I should be paying attention to for it?

 

One thing to note is that I got the parity drives after I had built the system. I transferred data to the array storage disks before they were installed. Not sure if that makes a difference.

 

Is anyone able to give me some advice on what I might need to look into? Not quite sure what other details to provide here that would be relevant. I'm planning to add one or two drives to the array every month to expand the storage capacity over time, but am a little worried that it might not be the best investment with this kind of performance.

 

Any advice on where I could go from here would be fantastic.

Edited June 10, 20206 yr by TensaZangetsu

4 minutes ago, TensaZangetsu said:

Any advice on where I could go from here would be fantastic.

Attach the diagnostics zip file intact to your next post in this thread.

I should have thought of that. My bad.

vectorsigma-diagnostics-20200610-2339.zip

Nothing jumping out, though you're using a RAID controller and those are not recommended, they can sometimes have unpredictable performance issues, cancel/pause the sync and run a diskspeed test to check disks are performing normally.

Interesting results, the parity drives actually seem to perform better in some places, despite them being an inferior model.

 

 

 

 

I seem to be getting by at the moment by pausing the parity sync for a couple of minutes at a time, then when it is resumed it goes 100 times faster for the next 5 gigs before slowing down again. Then repeat. Very tedious, but better than waiting 5 days.

 

 

It does look like I will need to get a non RAID card here, the other hardware is kind of wasted if disk performance is this undependable. Bit of a shame as that H730P card seems pretty powerful otherwise.

 

Any recommendations for what I should replace it with?

 

 

On a second look the problem might be the parity disks, those disks are SMR, and while SMR generally work great with Unraid I remember havinbg very bad performance with those specific models, most SMR disks behave normally during sequential writes, not those, they slow down a lot as soon as they fill the small PMR cache and need to write to the SMR zone, even for sequential writes.

SMR would explain it. The parity runs fine while the drives are still using their pseudo-CMR buffer, then once they move to SMR tracks the performance falls off a cliff.

 

If it were me, I'd replace them, but it depends on your use-case. Those drive probably have 10-20GB worth of buffer on them, so if you can get off this initial massive write, and the rest of your writes to the array aren't going to be very big (less than the buffer size) then the performance won't be too bad. That said, as Johnnie points out, it will be unpredictable.

Thank you very much. That makes a lot of sense, since I don't really have much trouble with the WD Reds, which aren't SMR.

 

I thought that having larger disks for parity would help with future proofing, in case I got larger storage drives later on, but WD Reds did not come in that capacity so I went with these WD Blues.

 

I think I'll just get the same model of WD Reds and see how that goes. Probably cheaper than getting a new card, and if it isn't the drives I can always add them to my storage array and work towards that solution.

 

Thanks again, I'll post back on how I get on when I swap out the drives.

Edited June 11, 20206 yr by TensaZangetsu

Double-check what drives you get. I assume you're aware of the news about WD shoving SMR everywhere they can manage it (including a bunch of Red drives) recently.

Yeah, not their finest moment, is it?

 

The 1TB WD Reds are the only ones at that capacity and form factor that are CMR, everything else in that form factor is either SMR or a lower capacity. I should have taken this into account when I bought the WD Blues, but I got excited with the thrill of the build and forgot to check.

I realised that I had enough space to clear out one of the WD Reds, so I took it out of the array and made it parity.

 

Made a world of a difference, definitely a 'THAT'S more like it' moment.

 

 

 

Thanks everyone for your help here!

You're welcome, and if you don't mind going to tag it solved.

I was just about to do that, thanks!