Pages

Monday, November 4, 2024

What is ESXi Core Dump Size?

ESXi host Purple Screen of Death (PSOD) happens when VMkernel experiences a critical failure. This can be due to hardware issues, driver problems, etc. During the PSOD event, the ESXi hypervisor captures a core dump to help diagnose the cause of the failure. Here’s what happens during this process.

Sunday, November 3, 2024

List devices

esxcli storage core device list

Get S.M.A.R.T information

[root@esx24:~] esxcli storage core device smart get -d t10.NVMe____KINGSTON_SNVS1000GB_____________________55FA224178B72600

[root@esx24:~] esxcli storage core device smart get -d eui.0000000001000000e4d25c0f232d5101


The output looks like this


VMware KB: https://knowledge.broadcom.com/external/article/314303/displaying-smart-data-for-nvme-devices-i.html

Saturday, November 2, 2024

Install and setup xfcw window manager in Centos 9

Install and setup xfcw window manager in Centos 9

sudo dnf -y groupinstall "Xfce"

Optional if multi-user console is used as default target. This would configure the system to boot into the graphical interface by default.
sudo systemctl set-default graphical.target

echo "xfce4-session" > $HOME/.xsession

chmod +x $HOME/.xsession

sudo reboot

Applications are defined in directory /usr/share/applications
Every application has its own definition file with file extension .desktop

For example you can use file chrome.desktop with the following content

[Desktop Entry]
Version=1.0
Name=Chrome
Comment=Google Chrome
Exec=/opt/google/chrome/chrome
Icon=/opt/google/chrome/product_logo_64.png
Terminal=false
Type=Application
Categories=Network;WebBrowser;


Thursday, October 31, 2024

pktcap-uw and tcpdump-uw

List VMs and their uplinks.

netdbg vswitch instance list

Capture DHCP traffic (udp 67, udp 67) n vmnic0 interface and send it to tcpdump to filter DHCP communication. 

pktcap-uw --uplink vmnic1 --capture UplinkRcvKernel,UplinkSndKernel -o - | tcpdump-uw -r - udp port 67 or udp port 68

14:45:46.375602 IP 0.0.0.0.bootpc > 255.255.255.255.bootps: BOOTP/DHCP, Request from 00:50:56:99:fe:6a (oui Unknown), length 300
14:45:46.376233 IP 192.168.4.5.bootps > 192.168.4.178.bootpc: BOOTP/DHCP, Reply, length 307


Filter TCP Open Connections

This is the tcpdump command to display attempts to open TCP connections (TCP SYN) from IP address 192.168.123.22

pktcap-uw --uplink vmnic4 --capture UplinkRcvKernel,UplinkSndKernel -o - | tcpdump-uw -r - 'src host 192.168.123.22 and tcp[tcpflags] & tcp-syn != 0 and tcp[tcpflags] & tcp-ack == 0'

Explanation:

  • -n → Do not resolve hostnames.
  • -i <interface> → Specify the network interface (e.g., eth0).
  • 'src host 192.168.123.22' → Filter packets from the source IP 192.168.123.22.
  • 'tcp[tcpflags] & tcp-syn != 0' → Match packets where the SYN flag is set.
  • 'tcp[tcpflags] & tcp-ack == 0' → Ensure the ACK flag is not set (to exclude SYN-ACK responses).

 

 

Wednesday, October 30, 2024

Colossus AI Supercluster with over 100,000 @NVIDIA H100 GPUs

Source: https://nvidianews.nvidia.com/news/spectrum-x-ethernet-networking-xai-colossus

Inside the @xai Colossus AI Supercluster with over 100,000 @NVIDIA H100 GPUs. If you want to see why the @Supermicro_SMCI liquid-cooled cluster is awesome, then check this one out.

O čem to video je?

100 000 GPU v datacentru

2 CPU and 8 GPU in 4U server chassis
8x server per rack
Takže 64 GPU per rack

1 563 racků v datacentru

Chlazení kapalinou. Liquid cooling.