Purpose
Utility tool for mapping of Vast clusters' internal network cnode/dnode.
Supports Ethernet and IB clusters.
Latest Script Version
Latest Version Number: v2.0
Latest Version Release Date: Feb 28th, 2024
Pre-requisites
1. All nodes can be accessed
2. switches are powered up
3. Supported switch types: Mellanox, Aruba, Cumulus
4. MGMT connectivity between cnodes <-> switch
How to Run the Script
vnetmap is available on Vast OS nodes
The latest version can be found attached to this article
We need to provide the Vast internal network mgmt IPs, ssh to a Vast CNode and run and run
cnodes_ips=$(clush -g cnodes echo | awk -F ':' '{print $1}' | paste -sd ',' -)
dnodes_ips=$(clush -g dnodes echo | awk -F ':' '{print $1}' | paste -sd ',' -)NOTE: Running with -discover flag will try to discover the nodes based on local node clush configuration file.
The IPs can be found on each node by looking for 69:m label on the mgmt bond
~:$ ip a | grep 69:m
inet 10.10.128.32/18 brd 10.10.191.255 scope global bond0.69:mexport cnodes_ips=`echo 10.10.128.{1..20} | sed 's/ /,/g'`
export dnodes_ips=`echo 10.10.128.{100..109} | sed 's/ /,/g'`
export SWITCH_IPS="10.255.255.253,10.255.255.252" <<<< Only needed for ETH, for IB it will be auto discoveredRun the script:
ETH:
python3 vnetmap.py -s $SWITCH_IPS \
-i $cnodes_ips,$dnodes_ips \
-u admin \
-p admin \
-k /home/vastdata/.ssh/id_rsa
IB:
python3 vnetmap.py -i $cnodes_ips,$dnodes_ips \
-k /home/vastdata/.ssh/id_rsa -ib
All options
$ python3 vnetmap.py -h
usage: vnetmap [-h] -s SWITCH_IPS -i HOST_IPS -u USER [-p PASSWORD]
[-k SSH_KEY] [-no-mtu] [-upload] [-subsystem] [-discover]
[-no-diag] [-d]
optional arguments:
-h, --help show this help message and exit
-s SWITCH_IPS, --switch-ips SWITCH_IPS
switch ips
-i HOST_IPS, --host-ips HOST_IPS
host ips
-u USER, -user USER user for switch
-p PASSWORD, -password PASSWORD
password for switch
-k SSH_KEY, --ssh-key SSH_KEY
ssh key to use
-no-mtu, --no-mtu-check
check MTU for each internal IP\Interface
-upload, --upload-s3 upload mapping file to vast
-subsystem, --subsystem-breakdown
print output per subsystem - for large scale clusters
-discover, --discover-nodes
auto discover nodes based on clush config
-no-diag, --no-diag-network
creates a report to diagnos fabric issue,
using -no-diag disables this check
-d, --debug for the impatient..print every step.Download vnetmap.py
Log in to Download