Jump to content

BUG: soft lockup - CPU (proxmox node)

I have a server on the data center with proxmox running specs(its also in a cluster of other 10 proxmox)
 

CPU(s)-32 x 13th Gen Intel(R) Core(TM) i9-13900 (1 Socket)
Kernel Version-Linux 5.15.116-1-pve #1 SMP PVE 5.15.116-1 (2023-08-29T13:46Z)
PVE Manager Version -pve-manager/7.4-16/0f39f621


Ram - 62GB

Have 2 NVME Samsung SSD 2 TB each both in LVM

I'm running 2 vm with 16 core 30 GB RAM and 1TB storage each both vm running (vicibox-8.0.1 server for client)

it was running with no issue for more than 3 months but for the last 2 days every morning i check the node is getting offline and when i check the log last entry in the log is
 

Dec 28 02:11:45 ProxVE12 kernel: watchdog: BUG: soft lockup - CPU#31 stuck for 417s! [CPU 6/KVM:43483]


attached syslog file for the last crash last 3min its the node syslog i do not have access of the client vm user info for the VMs log

Im new to this i do not know how much and which data I need to share here for help so if anyone needs any more data just tell me how to get the data i will manage to share it here need help with this, please

Thank you

anothercrash.txt

Link to comment
Share on other sites

Link to post
Share on other sites

  • 4 weeks later...

@koushik401

Try only allocating 14 threads to each VM, and is it serously necessary to allot 32 GB of memory to each virtual machine?

 

Try setting the block scheduler from cfq/bfq to noop to remove the scheduler from the guests, (on each separately) when the host will also have its own disk scheduling.

 

echo noop > /sys/block/sda/queue/scheduler

 

So do that for all main block devices, in both virtual machines and it should improve storage performance.

: JRE #1914 Siddarth Kara

How bad is e-waste?  Listen to that Joe Rogan episode.

 

"Now you get what you want, but do you want more?
- Bob Marley, Rastaman Vibration album 1976

 

Windows 11 will just force business to "recycle" "obscolete" hardware.  Microsoft definitely isn't bothered by this at all, and seems to want hardware produced just a few years ago to be considered obsolete.  They have also not shown any interest nor has any other company in a similar financial position, to help increase tech recycling whatsoever.  Windows 12 might be cloud-based and be a monthly or yearly fee.

 

Software suggestions


Just get f.lux [Link removed due to forum rules] so your screen isn't bright white at night, a golden orange in place of stark 6500K bluish white.

released in 2008 and still being improved.

 

Dark Reader addon for webpages.  Pick any color you want for both background and text (background and foreground page elements).  Enable the preview mode on desktop for Firefox and Chrome addon, by clicking the dark reader addon settings, Choose dev tools amd click preview mode.

 

NoScript or EFF's privacy badger addons can block many scripts and websites that would load and track you, possibly halving page load time!

 

F-droid is a place to install open-source software for android, Antennapod, RethinkDNS, Fennec which is Firefox with about:config, lots of performance and other changes available, mozilla KB has a huge database of what most of the settings do.  Most software in the repository only requires Android 5 and 6!

 

I recommend firewall apps (blocks apps) and dns filters (redirect all dns requests on android, to your choice of dns, even if overridden).  RethinkDNS is my pick and I set it to use pi-hole, installed inside Ubuntu/Debian, which is inside Virtualbox, until I go to a website, nothing at all connects to any other server.  I also use NextDNS.io to do the same when away from home wi-fi or even cellular!  I can even tether from cellular to any device sharing via wi-fi, and block anything with dns set to NextDNS, regardless if the device allows changing dns.  This style of network filtration is being overridden by software updates on some devices, forcing a backup dns provuder, such as google dns, when built in dns requests are not connecting.  Without a complete firewall setup, dns redirection itself is no longer always effective.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×