Upgraded yesterday from 6.1.9 to 6.2.  Upgrade went well.  I followed the VM upgrade process successfully and all dockers had updates available which I updated.  All seem to be working without issue.  However, I now notice that the CPUs are never idle even when (as far as I know) there should be no activity on the array.  I have a Haswell i5-4590 CPU and one or two of the four CPU cores is almost always at high utilization.  This is not constant for a given core.  One CPU will jump to 90-100% load and then after just a couple of seconds it drops back down to 0%-2% and after 3-4 seconds at idle, another CPU load goes to 100%.  Sometimes it is two CPUs that jump up in load with one at, for example 66% and another at 34%.

 

The result of this is that my CPU temps rarely drops below 43 degrees whereas on 6.1.9 idle temp was 37 degrees without the random short-lived high CPU loads.

 

With 6.1.9, I had to add the append intel_pstate=disable initrd=/bzroot line to syslinux.cfg to disable CPU turbo mode or I saw similar random high utilization patterns.  The intel_pstate=disable argument remains in syslinux.cfg.  If I remove it, CPU load and temps go even higher.

 

I ran top to see what was using CPU and it appears that, at "idle" the biggest consumer of CPU cycles is the "find" process.  It jumps to the top of the process list and then disappears as the loads drop down and then jumps back up to the top with high CPU utilization.

 

Frankly I am not sure if this an unRAID 6.2 OS issue or if a docker update is to blame since both happened more or less at the same time.  As a test, I stopped my dockers one by one to see if the pattern changed.  It did not.  Installed dockers are CrashPlan, DelugeVPN, Dolphin, Guacamole, Handbrake, OpenVPN-AS and Plex.

 

This has been going on since I upgraded to 6.2 more than 24 hours ago.

 

Diagnostics attached should they be useful.

 

 

 

medianas-diagnostics-20160918-1524.zip

Screenshots that illustrate issue:

 

 

What plugins do you have installed?  I know the old CachDirs plugin used find to do it's magic and I think the v6 Dynamix might too.  But the "find" command is a common command that could be used by lots of things.  So I know you tried your dockers but did you try disabling any plugins too?

Installed Plugins (all current as of this post):

-Community Applications

-Dynamix

  -Active Streams

  -Cache Directories

  -Local Master

  -S3 Sleep (Not Used)

  -System Buttons

  -System Information

  -System Statistics

  -System Temperature

  -webGUI

-Fix Common Problems

-Nerd Tools

-Preclear Disks

-Unassigned Devices

-unRAID Server OS

 

I'll try disabling one by one those I can.

Installed Plugins (all current as of this post):

-Community Applications

-Dynamix

  -Active Streams

  -Cache Directories

  -Local Master

  -S3 Sleep (Not Used)

  -System Buttons

  -System Information

  -System Statistics

  -System Temperature

  -webGUI

-Fix Common Problems

-Nerd Tools

-Preclear Disks

-Unassigned Devices

-unRAID Server OS

 

I'll try disabling one by one those I can.

Start with -Cache Directories.

First step was to boot in Safe Mode which loads no plugins.  In safe mode the problem is not present as seen the following screen grab (all CPUs remain at less than 5% load while in safe mode):

 

 

Now to try booting with plugins enabled and removing them one by one.  I will start with Cache Directories.

The culprit is the Dynamix Cache Directories plugin.

 

With it removed, "find" never shows up in the top processes list and CPU load is consistently down in the 0-10% on all CPUs.  As soon as it is reinstalled, back comes "find" with high CPU usage in the top list and one or more random CPUs go into brief high-load states (up to 100%) with resulting increased CPU temps.

 

It's a shame as cache directories has really helped when browsing the over 100,000 photos photos on the server stored in dozens of directories.  Without it, browsing directory contents and thumbnail generation is noticeably slower in directories with a large number of files.  It looks like I have a choice between increased CPU usage and heat or faster folder browsing.

 

I do not recall seeing this issue with whatever version of the cache dirs plugin I had installed with unRAID 6.1.9, but, perhaps I was just not paying attention to the details as much as I did after an unRAID OS upgrade and the issue was there earlier.

 

The version of the plugin is 2.1.1 according to the cache directories plugin config page or 2016.08.26 according to the plugin page.

 

This thread should be moved to the Plugins (V6) forum as it is not an unRAID OS issue.

Are you running the VM's when this happens?    If that is the case how much RAM are you leaving unused?  Leaving too little free could easily result in the Cache Directories plugin causing the effect you mention as it effectively 'thrashes' trying to keep directory entries in RAM without sufficient free RAM to do it effectively.

The server has 16 GB RAM.  I only have one VM (Windows 10) which is allocated 2GB initial RAM and up to 4GB max.  It is not currently running so should not be using any RAM.  According to dashboard page, memory usage is at 22%.  Right now, the only activity is Cache Directories doing its thing and causing fluctuating high CPU loads.

When I updated from 6.1.9 to 6.2, I also saw very high CPU values at first, but never isolated which component was responsible.  I went looking through the system to make sure everything was updated, uninstalled a few things I didn't need, and discovered the tools of the NerdTools were not updated and did that (just retoggle them).  There may have been something else too that I did, but once these were updated, my CPU's returned to very normal and low values.  I wish I knew exactly which change made the difference, but I don't.

 

My CacheDirs uses 0% to 4% CPU now, and I believe that's normal for some, high for many.  I only have an older AMD, nothing fancy.  There's no way yours should be doing 60%.

Thanks RobJ.  Something unusual is going on.  I checked the Nerd Tools and a couple needed updates, but, that did not appear to make a difference.  I currently have all Nerd Tools turned off and am seeing the same behavior.  I also think it is unusual that Cache Dirs would constantly need so much CPU.  Right now, the only thing I know for certain is that removing cache dirs results in a return to low CPU loads and system temps without the wild fluctuations.  I'll just have to keep digging when I have time.

I excluded several folders from Cache Directories that have few entries leaving only those with the most sub-folders/files and it seems to have helped immensely.

 

EDIT:  By excluding folders one by one, I was able to determine that caching the appdata and backups shares (either one or both), results in very high CPU loads.  With both excluded, CPU loads at "idle" fluctuate between 0-10% on all CPUs.  Since appdata is on the SSD cache drive, caching is not necessary as it is always spun up and not caching backups is no big deal.