As you can see, ‘data-collector’ is eating CPU cycles. Logged in via SSH, no browsers active. But, when I do try to view dashboard, etc - it is painfully slow to display.
System is running the latest Rockstor version: 3.9.2-49
4 cores, stand alone - not a VM. About 4TB of share.
Second image is RAM usage.
From first image, you can see I’ve recently rebooted the thing. It also appears it’s been running for some time.
Looking for ideas and guidance to troubleshoot, diagnose and hopefully fix this.
Yes, we have an issue with a newer library that was pulled in when 3.9.2-49 was rolled out.
Looks to mainly be a problem with webkit / Chrome browsers. Firefox looks to be OK though.
Please see the following thread where it looks to have now been sorted:
I’ve now rebuild all our rpms and they are in final testing. Hoping to roll the stable channel instance out tomorrow as 3.9.2-50. If all is well that should sort this out. Very much looks like it from what we’ve tested so far. Worth reading whole of above thread as there were other reports of the data-collector high CPU usage. I only managed to trigger it once myself and wasn’t able to reproduce but the 10-30 second pauses in chrome were reliable reproducible.
Ended up just pinning a library version that we nearly pined months ago but it ended up being ‘fixed’ upstream back then; only to revisit us in this last release.
Hope that helps and thanks for the report. Let us know if 3.9.2-50, once it’s published, sorts things for you. Ironically it would have been one of the fastest stable releases for ages Web-UI wise, and does seem to work for most in Firefox but alas; bit by bit. I’ll most likely announce the rpm availability in that thread. And do a quick post to update the following changelog thread:
That explains why I couldn’t see things in Chrome and had to switch to Firefox.
However, that’s not my problem .
See the CPU utilization for ‘data-collector’ in my screenshot there? Yea - that’s the problem. No browser involved, nothing even looking at any web interface. And, even in Firefox - the screens did finally display and looked fine - but they took over 2 minutes to load. No problem here to use FF instead of Chrome.
Problem appears to be data-collector hogging the CPU and slowing everything else down. This box has plenty of iron to handle things, yet the CPU is still bogged.
HOWEVER!!
I just did another look see and data-collector is NOT running at all. CPU utilization is back down to normal and she’s running fine. Even after firing up Firefox and poking about - all normal.
I guess this is an intermittent task that hogs the CPU when it runs, eh?
