Added nv-mlnxipcfg.py

README.md

@@ -9,4 +9,93 @@ pages and provided as a reference for recipe implementation
 It is recommended for use during bring-up and that you implement only
 required components for deployment in production environments
 
-use "-h" for more details
+use "-h" for more details
+
+
+# nv-mlnxipcfg.py
+
+nv-mlnxipcfg.py configures and sets persistency behavior for IP addresses, IP rules, IP route, and ARP settings
+
+Requires: Python 3.x and python3-click
+
+nv-mlnxipcfg.py --help
+
+Usage: nv-mlnxipcfg.py [OPTIONS]
+
+Options:
+  -v, --verbose       verbose logging
+  -i, --ipaddr TEXT   Starting IP address with netmask (for example:
+                      192.168.1.1/24 or 2001::66/64)  [required]
+  -f, --flush         Flush IP addresses before adding new ones
+  -r, --dryrun        Dry run. Do not actually assign IP addresses or add
+                      netplan cfg
+  -d, --devices TEXT  comma separated device list (for example:
+                      enp225s0f0,enp225s0f1) if '-d' not provided, tool will
+                      configure all found Ethernet devices
+  --help              Show this message and exit.
+```
+Debian/Ubuntu Based OS
+======================
+1. A required configuration of a starting IP address (IPv4 or IPv6) with netmask (i.e. 192.168.1.1/24) is provided.
+2. It then looks at all the Infiniband interfaces it can find (under /sys/class/infiniband/*)
+3. It then checks to make sure they are "Ethernet" link_type by checking /sys/class/infiniband/{}/ports/*/link_layer
+4. If the flush option is provided (-f or --flush) it will remove all the IP address first on the interface (ip addr flush ...)
+5. Now for all the Ethernet type interfaces, it adds an IP address starting with the one provided.
+6. It will then run 3 ip route and one ip rule commands using the source IP address and a default gateway (Note: I assume the default gateway is one IP address less then the broadcast address for the network provided):
+  a. ip route add 0.0.0.0/1 via {} dev {} table {} proto static metric {}
+  b. ip route add {} dev {} table {} proto static scope link src {} metric {}
+  c. ip route add 128.0.0.0/1 via {} dev {} table {} proto static metric {}
+  d. ip rule add from {} table {} priority 32761
+7. It will then run the sysctl command and set the following for each interface
+  a. net.ipv4.conf.{}.arp_accept=1
+  b. net.ipv4.conf.{}.arp_announce = 1
+  c. net.ipv4.conf.{}.arp_filter = 0
+  d. net.ipv4.conf.{}.rp_filter = 2
+  e. net.ipv4.conf.{}.arp_ignore = 1
+8. After that, it will create a netplan configuration with the IP addresses, routes, and route-policy as shown in the docs I was provided with the same config used in the ip addr add , ip route and ip rule commands. The file it will write to maintain persistence is called /etc/netplan/55-nvidia-autoconfig.yaml.
+9. The sysctl configuration file is constructed with the same values for each interface for the 5 ARP and RP settings and written to /etc/sysctl.d/55-nvidia-arpdefaults.conf
+
+RHEL Based OS
+=============
+To support for RHEL and CENTOS servers, the ifcfg and route scripts are written to
+
+      /etc/sysconfig/network-scripts/
+
+These are needed for persistence upon reboot.
+
+In addition to the previous configs (ip addr add, ip route, ip rule, sysctl ARP settings) for RHEL servers,
+this script adds an ifcfg and route config file for each interface on RHEL and CENTOS servers instead of a NETPLAN config file.
+
+For example, for the single interface  enp225s0f0  with address 66.66.66.66/24:
+
+[lab@cl1-fair-01 ~]$ cat  /etc/sysconfig/network-scripts/ifcfg-enp225s0f0
+BOOTPROTO=none
+NAME=enp225s0f0
+DEVICE=enp225s0f0
+ONBOOT=yes
+IPADDR=66.66.66.66
+PREFIX=24
+DEFROUTE=yes
+GATEWAY=66.66.66.254
+ROUTING_RULE="priority 32761 from 66.66.66.66 table 101"
+IPV4_FAILURE_FATAL=no
+IPV6INIT=yes
+IPV6_AUTOCONF=yes
+IPV6_DEFROUTE=yes
+IPV6_FAILURE_FATAL=no
+
+[lab@cl1-fair-01 ~]$ cat  /etc/sysconfig/network-scripts/route-enp225s0f0
+ADDRESS0=0.0.0.0
+NETMASK0=128.0.0.0
+GATEWAY0=66.66.66.254
+METRIC0=101
+OPTIONS0="table 101"
+ADDRESS1=128.0.0.0
+NETMASK1=128.0.0.0
+GATEWAY1=66.66.66.254
+METRIC1=101
+OPTIONS1="table 101"
+ADDRESS2=66.66.66.0
+NETMASK2=255.255.255.0
+METRIC2=101
+OPTIONS2="onlink src 66.66.66.66 table 101"