FreeNAS 11 Constant High CPU from Python3.6 process

[Ericloewe]

2) Which web browser are you using? Firefox on my system causes more usage wheras IE or Chrome uses less.

On the server?

 

[joeschmuck]

I still think your numbers are fine.

What is FreeNAS 11 accomplishing that takes 10 times the CPU work to accomplish over 9.3?

Not sure that is a far comparison. FreeNAS 11 is vastly different from FreeNAS 9. It's build on FreeBSD 11 vice 9 for one reason.

Also, just refreshing my browser often in the GUI I now have values of 1.50, 1.54, 1.11. Just showing you that the browser alone can affect the load.
Now I've left it alone for a few minutes and it changed again to: 0.25, 0.89, 1.03

The uptime command means: This is the average number of jobs in the run queue over 1 minute, 5 minutes, and 15 minutes.

I just think you are troubleshooting someting which isn't a problem.

My system used to show numbers like 0.01. My main concern is heat. Since this is a silent system, this CPU use is causing some high temperatures in the case that used to be cold to the touch at all times.

I find it very hard to believe that the low values that you are displaying are causing you high CPU temps.
Just to let you know, I kicked off a SCRUB and this barely shows up on the loads however it definately shows up on the CPU Usage Chart. Since you are experiencing CPU temperature issues I would suggest you turn your troubleshooting towards the CPU. The command top can show you who is using your CPU time. For my SCRUB it turns out that "system" is using most of my CPU time using up to ~20% of it up. And my load averages are all below 0.9x right now.

I'm not going to tell you that it will be easy to figure out what is driving up your CPU temps, it can be a royal pain. Using command ps may help but it may be chasing your tail too.

Can you post your CPU Utilization Chart?

 

[joeschmuck]

On the server?

Nope, on my main computer. FireFox just sucks in this respect. I dont' know why, I just accept it. Heck, if FireFox isn't goint to support Java anymore I might as well switch to Chrome.

 

[Ericloewe]

Nope, on my main computer.

But then it's not really related to the server's abnormal CPU usage.

 

[joeschmuck]

But then it's not really related to the server's abnormal CPU usage.

I honestly do not know why the server shows a higher load when I use FireFox over IE or Chrome but it does show up on the server under the load. I gave up on HTML and web browser programming decades ago.

 

[Ericloewe]

So it does show up on the server. I guess that's from firefox running some script at a higher frequency (possibly too high).

 

[FreeNASBob]

I am using SSH to check uptime and top, etc. I manually applied the patch to autosnap.py and it has improved the load. I now see numbers more like 0.22, 0.20, 0.21.

What's interesting to me is looking at the memory usage before patching autosnap.py and rebooting vs. after. Can they be connected?

 

[danb35]

What's interesting to me is looking at the memory usage before patching autosnap.py and rebooting vs. after. Can they be connected?

Unlikely--the system will start using RAM as cache over time, but that won't have happened to a significant degree in 15 minutes.

 

[FreeNASBob]

Almost 8 hours later and RAM usage still looks like the screen capture above ... holding at 814MB.

 

[MrToddsFriends]

I am using SSH to check uptime and top, etc. I manually applied the patch to autosnap.py and it has improved the load. I now see numbers more like 0.22, 0.20, 0.21.

Did you measure the power consumption of your system before and after this change? If so, how? Is there a substantial difference?

One more thing: If you are seeing high CPU temperatures (for example in IPMI) while the CPU heatsink stays cool there might be a heat transfer problem as seen in some copies of Asrock C2x50D4I boards, for example here:
https://forums.freenas.org/index.ph...pu-temperature-issue.41528/page-3#post-267639
The whole thread has 5 pages, maybe reading from start is necessary to understand what's going on.

What's interesting to me is looking at the memory usage before patching autosnap.py and rebooting vs. after. Can they be connected?

Unlikely. "Using RAM as cache" is depending on usage not on uptime.

 

[MrToddsFriends]

I am using SSH to check uptime and top, etc. I manually applied the patch to autosnap.py and it has improved the load. I now see numbers more like 0.22, 0.20, 0.21.

BTW: Thanks for trying this out and reporting the effect on system load you were seeing. "The patch to autosnap.py" is the one discussed in Feature #18428 mentioned earlier in this thread, correct?

Don't forget that patches you apply to the system will be overwritten with the next update.

 

[joeschmuck]

Almost 8 hours later and RAM usage still looks like the screen capture above ... holding at 814MB.

If you have done nothing with the system then odds are the RAM usage will not grow very fast. As previously stated, it's likely the cache that you freed up with the reboot. The cache will grow as you use the system.

 

[FreeNASBob]

The RAM usage has still not changed significantly after a day of use. Either something was broken before, or that autosnap.py really had an effect.

 

[MrToddsFriends]

The RAM usage has still not changed significantly after a day of use. Either something was broken before, or that autosnap.py really had an effect.

Are snapshots and replications (assuming there are some configured in your setup) still executed in your current patched software state?

 

[FreeNASBob]

No replications, but 4 snapshots run daily and they are showing up in the snapshots UI.

 

[Chris Moore]

Almost 8 hours later and RAM usage still looks like the screen capture above ... holding at 814MB.

Is any read / write actually happening. ARC doesn't fill until there is network communication to the storage.

Sent from my SAMSUNG-SGH-I537 using Tapatalk

 

[FreeNASBob]

Is any read / write actually happening. ARC doesn't fill until there is network communication to the storage.

Snapshots, a scrub, and using Plex.

I just checked again and noticed something strange. The ARC size stayed in that 800 MB range for 35 hours, then suddenly at 2:04 AM today it increased 600% in size to over 5G even though there are no scrubs, snapshots, or anything else scheduled anywhere near that time and no users were connected for hours before or after.

 

[Stux]

no replications triggered at 2am?

Something is kicking off at 2AM.

 

[FreeNASBob]

no replications triggered at 2am?

Something is kicking off at 2AM.

I don't have anything scheduled for 2AM. The nearest thing is a snapshot at 12:00 am. I'm trying to find something in the logs, but nothing yet. All I can tell from the graphing is that it was pretty much a 100% read operation that lasted about a minute. I can see a spike in CPU at the same time, but no increased network activity. I was thinking maybe a SMART test, but those are scheduled for Sundays at 3:00 am.

 

[FreeNASBob]

I checked crontab -l and browsed /etc/cron.d and all are empty or don't exist. I tried from root as well as from a shell in the Plex jail.