Hello hoping for a little guidance please,

 

My Unraid is good and works well. apart from when I copy large files between my shares, which seems to make my Plex unresponsive (plex sometimes complains media files don't exist for Movies folders and hangs until files transfer is complete)

 

I have a cache pool of 4 SSDs, all on Sata 3-6Gbps (2 of Mobo and 2 on StarTech PCI-E Sata3 -6Gbps controller(s) - 2 StarTech Sata3 controllers, so SSD's not sharing PCI-E.. just in case)

 

OCZ-VERTEX3_OCZ-66HCYN64A85WJ3O9 - 120 GB (sdd)

OCZ-VERTEX3_OCZ-975SC76BI6T80JUW - 120 GB (sdi)

KINGSTON_SV300S37A120G_50026B766700DF19 - 120 GB (sdh)

SanDisk_SDSSDP256G_131533401507 - 256 GB (sdc)

 

Shares are WD Cavier Reds (connected via Sata-2 3Gbps ports on Mobo)

WDC_WD30EFRX-68EUZN0_WD-WMC4N1499423 - 3 TB (sdf)

WDC_WD30EFRX-68EUZN0_WD-WMC4N1499423 - 3 TB (sdf)

 

Parity  (connected via Sata-2 3Gbps ports on Mobo)

WDC_WD30EZRX-00MMMB0_WD-WCAWZ1902298 - 3 TB (sdg)

 

Could it be I.O issues with Unraid managing 4 SSDs in Cache? Would I be better buying & migrating cache from current  4 SSD's to 2 x 256Gb SSD's?

 

a bit more info:

 

if I'm transferring files from PC (PC has Samsung Evo 850 pro SSD) to UnRaid share with Cache Share

I've noticed on transferring 10Gb movie starts at 130Mb for 10 seconds, then drops down to 50-60Mb for a constant speed)

But on transferring a 700Mb movie I get between 115 to 130Mb, so what i would expect from Cache SSD

 

if I'm transferring 10Gb movies between shares that have been moved off cache to Shares on WD Reds via my PC, I get about a constant 53Mb transfer

 

if I'm transferring 10Gb movies between UnRaid shares that have been moved off cache to Shares on WD Reds via Windows 10 Vm on Unraid (Win10 O.S on UnRaid SSD cache), initially starts at 130Mb, then drops after 10 (ish )second like a cliff down to 1 or 2mb at times, then after a few seconds back up tp 80Mb, then after a few seconds down to 10Mb.. which makes me think I.O issue between shares will copy from WD Red to SSD cache (I've checked) &  (yes copying this 10Gb movie with changed file name)

 

I started to think is this an issue with WD reds, even though i can download movies located on Shares on WD Red to pc at 80mbps and works wekk). But just in case, as I have a USB-3 WD drive connected to Windows 10VM on UnRaid, on transferring renamed 10Gb movie from USB 3 HDD (which isnt on a UnRaid share as only associated to Windows 10 VM, so doesn't use UnRaid Cache either) Starts at about 80Mb for 15 ish seconds, then drops between 3mb to 50Mb (ish) and plex although slight more responsive, still feels much slower than usual.. much slower!) after transfer Plex is snappy again

 

I've added my log data whilst I was completing the above tasks... any thoughts / guidance appreciated.. Please.

 

Cheers, Julian

 

 

 

 

 

 

UnRaid_21st_Jan_syslog.txt

Assuming that your share is set to cache "yes"?  If so assuming your cache isn't full it should be transferring at 70-80MB/s easily likely higher.

Post your diagnostcs zip (from the web GUI: Tools -> Diagnostics).

Assuming that your share is set to cache "yes"?  If so assuming your cache isn't full it should be transferring at 70-80MB/s easily likely higher.

 

Hello, yes cache is set to yes an there is 200Gb of available cache. I've attached pdf showing share settings and available cache

UnRaid_share_settings_and_available_cache.pdf

Post your diagnostcs zip (from the web GUI: Tools -> Diagnostics).

 

Thank you John. I've uploaded files as suggested. appreciate help!

tower-diagnostics-20170121-1450.zip

There's no information in the diagnostics zip - just empty folders.

 

Can you telnet or ssh into your server and type

 

diagnostics

 

at the prompt? The zip file should be located in the logs folder of your boot USB device. If that doesn't work you'll have to reboot and repeat the action that caused the problem and then get the diagnostics.

 

Havent SSHed into UnRaid (but I have when I was playing with Pi, so can again (ust need to remember how.. sorry lacking unix knowledge a bit). but hoping this re-download version works better, as have more in it this time?

tower-diagnostics-20170121-1511.zip

There's no information in the diagnostics zip - just empty folders.

 

Can you telnet or ssh into your server and type

 

diagnostics

 

at the prompt? The zip file should be located in the logs folder of your boot USB device. If that doesn't work you'll have to reboot and repeat the action that caused the problem and then get the diagnostics.

 

John, I havent SSHed into UnRaid (but I have when I was playing with Pi, so can again (I just need to remember how.. sorry lacking unix knowledge a bit). but hoping this re-download version works better, as havs more data in it this time?

tower-diagnostics-20170121-1511.zip

Yes, that's better. I'll get back if I spot anything.

I cant see anything dodgy looking through the logs, does this look a bug with using 4 SSD's in write cache? (If so i was going to raise a bug investigation with UnRaid engineering).

Yes, that's better. I'll get back if I spot anything.

 

Thanks John, I've also asked a guru at work to have a scan and we (he) can't find anything unusual in the logs, In your opinion does this look like a bug with using 4 SSD's in write cache (opposed to just 2)? (If so, i was going to raise a bug investigation with UnRaid engineering, as there isnt a suggested max limit for how many drives can be in a write cache in UnRaid Cache documentation).

I don't see anything out of the ordinary either.

 

Tell me more about your cache configuration. I see you have three 120 GB SSDs and one 256 GB SSD. Is that in fact the case, since in your OP you mention one SSD (sdi) twice? How is BTRFS configured with them? RAID 1 makes most sense to me. There's a capacity calculator here where you can experiment with different configurations before deciding.

 

It might be worth researching your particular models of SSD. Some manufacturers are getting sneaky and use slow NAND chips for the majority of the capacity and fast ones for the rest, so they appear fast at first but then slow down on large writes.

 

Do you trim them regularly?

 

I don't see anything out of the ordinary either.

 

Tell me more about your cache configuration. I see you have three 120 GB SSDs and one 256 GB SSD. Is that in fact the case, since in your OP you mention one SSD (sdi) twice? How is BTRFS configured with them? RAID 1 makes most sense to me. There's a capacity calculator here where you can experiment with different configurations before deciding.

 

It might be worth researching your particular models of SSD. Some manufacturers are getting sneaky and use slow NAND chips for the majority of the capacity and fast ones for the rest, so they appear fast at first but then slow down on large writes.

 

Do you trim them regularly?

 

John,

 

Good point about the Nand, I'll investigate this.

 

I do have 4 SSD's for cache (2 x OCZ's, 1 x SanDisck and 1 x Kingston),  I stupidly copied the same OCZ data twice for the same drive, opposed to copying Kingston ssd data (I've modified earlier post to reflect reality) I've uploaded the full data re the SSD caches and all the settings in pdfs attached here ( I had to do 2 replies to attach full settings pdfs as attachments exceeded 320kb)

 

I haven't setup up TRIM, I thought this was in UnRaid v6.x my default, so I assumed it was setup. I can see lots of instructions for enabling TRIM, any easy instructions as there are varying guidance over the last year between different UnRaid O.S versions (I'm trying to avoid coding/SSH if poss) you can recommended please?

Tower_Device_Cache_0.pdf

Tower_Device_Cache_2.pdf

I don't see anything out of the ordinary either.

 

Tell me more about your cache configuration. I see you have three 120 GB SSDs and one 256 GB SSD. Is that in fact the case, since in your OP you mention one SSD (sdi) twice? How is BTRFS configured with them? RAID 1 makes most sense to me. There's a capacity calculator here where you can experiment with different configurations before deciding.

 

It might be worth researching your particular models of SSD. Some manufacturers are getting sneaky and use slow NAND chips for the majority of the capacity and fast ones for the rest, so they appear fast at first but then slow down on large writes.

 

Do you trim them regularly?

 

John,

 

Good point about the Nand, I'll investigate this.

 

I do have 4 SSD's for cache (2 x OCZ's, 1 x SanDisck and 1 x Kingston), I stupidly copied the same OCZ data twice for the same drive, opposed to copying Kingston ssd data (I've modified earlier post to reflect reality) and I've uploaded the full data re the SSD caches and all the settings in pdfs attached here ( I had to do 2 replies to attach full settings pdfs as attachments exceeded 320kb)

 

I haven't setup up TRIM, I thought this was in UnRaid v6.x my default, so I assumed it was setup. I can see lots of instructions for enabling TRIM, any easy instructions as there are varying guidance over the last year between different UnRaid O.S versions (I'm trying to avoid coding/SSH if poss) you can recommended please?

 

Data for cache disks 3 & 4 attached here

Tower_Device_Cache_3.pdf

Tower_Device_Cache_4.pdf

Install bonienl's SSD Trim Plugin from here: https://lime-technology.com/forum/index.php?topic=36543.0 and see if it makes a difference.

 

The page for Cache0 has all the information about BTRFS. I see you're using the default RAID1 configuration, which makes sense with your three 120 GBs and one 256 GB.

Install bonienl's SSD Trim Plugin from here: https://lime-technology.com/forum/index.php?topic=36543.0 and see if it makes a difference.

 

 

Thanks John. I tried to install plugin, but I received a SSL failure error inside unraid Plugin tab, I've asked that forum for help separately.

 

On the other note. I've checked my OCZ's and they are awful with compressed media (aka Video files), see notes from http://ssd.userbenchmark.com/

 

'The 120GB OCZ Vertex 3 has good read speeds but lags the group leaders by 50% at writing. This is explained by its Sandforce 2281 controller which is infamous for stability issues and relatively slow incompressible data write speeds. Small file 4K peak and average read/write speeds clocked in at 31/103 and 23/61 MB/s which matches the group leaders, a superb result considering the Vertex 3's age. Overall the 120GB Vertex 3 has an effective speed of 383 MB/s which is only 16% lower than the group leaders which averaged 444 MB/s. The Vertex is a capable drive let down only by its relatively slow peak sequential write speed of 157 MB/s'

 

So I've decided to sell all my 4 SSD's on ebay and buy 2 x Crucial 550 256's. But before I do I will try sort out Trim just for giggles so see if that helps performance at all and report back. Plus Ill report back on comparable performance on Crucial 550's as I'm sure others will have lots of compressed videos on their Shares also.

 

btw Really appreciate your help John..

That looks like a corrupted download. There have been problems with GitHub's servers recently. Try again later is my suggestion.

 

That looks like a corrupted download. There have been problems with GitHub's servers recently. Try again later is my suggestion.

 

John I had the wrong time on UnRaid 6.2.4. (which keeps losing time) I edited date and time settings to pool.ntp.org, clicked apply and , the clicked done and time was correct and was able to download ok, without SSL error

Ah, OK. That makes sense.

 

The plugin doesn't appear to do very much but it adds a setting in Settings -> Scheduler for when and how often it is to run.

 

Ah, OK. That makes sense.

 

The plugin doesn't appear to do very much but it adds a setting in Settings -> Scheduler for when and how often it is to run.

 

John, my issues are now fixed! :-) With the Trimming you suggested. file transfers (with original 4 x SSD's) between shares is about 115MB/seconds and Plex is quick. Appreciate your guidance and patience with the newby Muppet i am.

Not totally sure this applies to your board since I can't determine the stepping from the syslog:

http://www.anandtech.com/show/4143/the-source-of-intels-cougar-point-sata-bug

 

TL;DR there was a bug in the PCH SATA controller that caused performance degradation.

 

Not totally sure this applies to your board since I can't determine the stepping from the syslog:

http://www.anandtech.com/show/4143/the-source-of-intels-cougar-point-sata-bug

 

Hello, my issues are now fixed! :-) I enabled TRIM plugin and now with original 4 x SSD's I'm getting about 115MB/seconds with coping between shares and Plex is quick. Sata cards and works ok. Appreciate the guidance. Cheers!