Updated from 6.1.9. All working fine now.

 

VMs worked fine after I'd been in to each one and changed the bridge br0 from virbr0.

 

I was prompted that the docker image should be recreated due to a bug in the betas. I did so.

I needed to change the appdata path from a user path to a device path.

Then I needed to recreate my docker templates. It had saved all the templates so it only took a few minutes for the 6 or so dockers.

 

I've also reformatted my cache drive as BTRFS and have moved all the appdata and vms to it. These were on an "unassigned" SSD.

 

I'll add a 2nd drive to the cache pool in a few days.

I am experiencing very poor performance copying data from my unRAID server (RC5) to my mac, specifically an external enclosure connected to my mac via thunderbolt. It used to take a few seconds to copy a 1GB file, now my mac is telling me 23 min. Also, if I want to copy a larger file, say an 8GB file, it used to take maybe 1 minute or less, now my mac is telling me over an hour. I did not have this problem under RC3. This is my 100TB server, which is connected to the network via a gigabit connection, nothing else has changed other than the update to RC5.

 

I just tried copying (from my Mac) an 8.79GB file from my unRAID server to one of my NAS's and it too is telling me its going to take 2hrs, something is not right.

 

In this current state, my 100TB unRAID server is useless to me as a NAS right now with these crazy slow copying speeds.

Gigabit is about 120MB/s at maximum throughput on large files.  Significantly less with smaller files.  An 8.7GB file should take about 9.5 minutes to READ off your UnRAID box.  Probably double that to write assuming you're not using a SSD cache.

 

1GB file would NEVER have taken "a few seconds", it's just over a minute best case scenario.

 

Gigabit is about 120MB/s at maximum throughput on large files.  Significantly less with smaller files.  An 8.7GB file should take about 9.5 minutes to READ off your UnRAID box.  Probably double that to write assuming you're not using a SSD cache.

 

1GB file would NEVER have taken "a few seconds", it's just over a minute best case scenario.

 

Something doesn't seem right with the math there.  If we're just talking about maximum gigabit network throughput (120MB/s):

 

1GB would transfer in ~9 seconds

8.7GB would transfer in ~73 seconds

I just tried copying a 1.03GB file from my mac to one of my other unRAID servers currently running 6.18, it is also connected via gigabit, it took 5 seconds. I tried copying it from my 100TB unRAID server to the same destination, the file copy dialog box said 11 minutes.

I just tried copying a 1.03GB file from my mac to one of my other unRAID servers currently running 6.18, it is also connected via gigabit, it took 5 seconds. I tried copying it from my 100TB unRAID server to the same destination, the file copy dialog box said 11 minutes.

Are you sure you don't have a networking issue? It is not uncommon for 1G ethernet to degrade to 100M.

No networking issues, all other computer on lan are functioning as expected with expected speeds and performance. The only change to this server was the upgrade yesterday to RC5 from RC3, I may have said I was running RC4, I was not, I went from RC3 to RC5.

No networking issues, all other computer on lan are functioning as expected with expected speeds and performance. The only change to this server was the upgrade yesterday to RC5 from RC3, I may have said I was running RC4, I was not, I went from RC3 to RC5.

Can you provide evidence that server has not degraded to 100M ethernet?

How would I go about doing that?

I just tried copying a 1.03GB file from my mac to one of my other unRAID servers currently running 6.18, it is also connected via gigabit, it took 5 seconds. I tried copying it from my 100TB unRAID server to the same destination, the file copy dialog box said 11 minutes.

 

Are there other client machines (besides your mac) on your network that you could test file transfer speeds to your 100TB unRAID server?

 

Would it be possible to try RC4 temporarily on your 100TB unRAID server just to verify network transfers speeds?  If RC4 is still slow, try RC3 to verify network transfers are at their usual gigabit speed.  This would help narrow down when the issue started happening.

 

 

Are you suggesting I roll back to RC3 or RC4 from RC5? How do I do that?

 

 

How would I go about doing that?

How do you know there is no networking issue if you don't have evidence?

 

Post a diagnostic.

 

I have no evidence to the contrary, everything (other than the file copying performance issue) appears to be working fine from a network standpoint. I just rebooted the switch that the 100TB server is connected to and now when I try to copy an 8.79GB file from the 100TB unRAID server to my mac it says 3hrs, better than 11 before.

 

FYI there are two network adapters in my server, only one is active at the moment.

 

Diagnostics attached.

tower-diagnostics-20160906-0900.zip

Are you suggesting I roll back to RC3 or RC4 from RC5? How do I do that?

On the flash drive, there is a folder called Previous.  copy bzroot and bzimage from inside it to the root of the flash drive and then reboot.

 

Thanks Squid, I may have to do that.

I have no evidence to the contrary, everything (other than the file copying performance issue) appears to be working fine from a network standpoint. I just rebooted the switch that the 100TB server is connected to and now when I try to copy an 8.79GB file from the 100TB unRAID server to my mac it says 3hrs, better than 11 before.

 

FYI there are two network adapters in my server, only one is active at the moment.

 

Diagnostics attached.

eth0 and eth1 both say speed 1000. How are you using both of these?

 

Are you trying to copy from a user share or from a disk? If from a user share you might try it directly from disk, trying multiple disks and multiple controllers just to see if you get a different result from specific hardware.

I am only using eth0

 

I am trying to copy from my downloads share to my mac, or my NAS, both are responding with terrible performance.

 

I think I will try rolling back to RC3 and see what it is like.

Ok so I rolled back to RC3, and I am now copying that same 8.79GB file to the same destination on my Mac and it just finished in a little over two minutes. So something is definitely fishy in RC5 with respect to how it applies to my system, not sure whether its drivers for the Intel nic in my system or what.

Gigabit is about 120MB/s at maximum throughput on large files.  Significantly less with smaller files.  An 8.7GB file should take about 9.5 minutes to READ off your UnRAID box.  Probably double that to write assuming you're not using a SSD cache.

 

1GB file would NEVER have taken "a few seconds", it's just over a minute best case scenario.

 

Something doesn't seem right with the math there.  If we're just talking about maximum gigabit network throughput (120MB/s):

 

1GB would transfer in ~9 seconds

8.7GB would transfer in ~73 seconds

 

Arse.    I took the lazy way out and used a calculator, and put Mbps instead of MBps. 

I have updated to RC5 just now with no issues, thank you.

No issues to report after upgrading two servers, one native and one on esxi. Great work guys!

So much hair splitting on forums... 

 

Updated to RC5....works great as usual...thanks.

 

 

No issues. Updated 2 servers. Thanks LT!

- add experimental global share setting: "Tunable (enable Direct IO)"

 

Why are new experimental features being added in an RC?

 

This.  RCs are for bug fixes, not adding more shit to break.  Adding stuff in RC5 totally and utterly defeats the purpose of an RC.  RC should be a feature freeze then bug fix, nothing else.

 

You guys are making a mountain out of a molehill. The new option is disabled by default and has to be turned on manually from the share settings page. This new "feature" is new to unRAID, but not new to fuse and has been available in fuse for some time now. It's not like this is akin to dual parity or something as far as a feature goes. In addition, we had done testing with this far before the RC phase, we just hadn't included the tickbox for turning it on in the web GUI.

 

The main purpose for this new option is for 10gbps network transfers as Johnnie Black already saw. He jumped from 700MB/s to full 1GB/s as a result.

 

on both of my unraid boxes running RC5 and going to the share settings page, the setting on both is auto for this feature.

on, off and auto are available. I haven't previously touched this setting on either box, auto is not off.....

on both of my unraid boxes running RC5 and going to the share settings page, the setting on both is auto for this feature.

on, off and auto are available. I haven't previously touched this setting on either box, auto is not off.....

 

According to the help text:

 

Auto = off