Introduction
One of our requirements with one of our customers is to perform regular and routine failover tests. As the interface is not responsive to providing information about when service has failed from Primary to Secondary and back again, I have found [1] and modified this perl script to run on the SECONDARY NODE to show the interface address of one NIC every 5 seconds. I’ll also show how to slightly modify the script with different time delays and interface names. Please note, there may be much better ways of doing this. I needed something in a hurry, and this gave me what I needed. If you’ve got any better ideas, please drop me a note at jon@spriggs.org.uk or update this script yourself :)
Steps to perform
- SSH to the Secondary node.
- Check you’re not already primary with the command ifconfig em0 | grep inet this should return one line showing something like inet 1.2.3.4 netmask 0xffffff00 broadcast 1.2.3.255
- Please note the exact syntax of this command – perl is a tricky beast at the best of times, and if you don’t have your apostrophies and backticks in the right place, it won’t run right!
Type this command: perl -e ‘for (;1;) {print `ifconfig em0 | grep inet`; print `date`; sleep 5}’ and press return. This loops until you press Ctrl+C, showing the line, like I showed you before starting inet and then the next line shows a date and time – this is so you don’t go crazy and think the process has stopped…
- Perform your action to provoke fail-over [2], which may be to unplug an interface attached to the primary firewall, reboot the firewall or unplug a switch directly attached to the firewall. In response (and after approx 1 minute, based on your HA configuration) you should now see in the script’s output, it now shows two lines – as follows:
inet 1.2.3.4 netmask 0xffffff00 broadcast 1.2.3.255
inet 1.2.3.5 netmask 0xffffff00 broadcast 1.2.3.255 - Perform your failback and after 1 minute or so, it should revert to just the single line – 1.2.3.4 or equivelent for your network.
Breaking down the script
Here, I’ve broken down the command before with short blocks of information about each part of the command you’re running.
perl (which is a scripting language) -e (which means “evaluate the following string in quotes”) ‘ (that’s an apostrophy – the symbol on my keyboard at least, below the @ character) for (;1;) (note, those are normal brackets – shift+9 and shift+0, which, in combination with the for and ;1; bits, means loop around the following commands forever, or until the break command is sent) { (that’s a curled brace symbol – shift and the character next to the letter P) print (sends the following string) ` (backtick – the character to the left of the 1 key – which means execute the command between two of these symbols and return the result) ifconfig em0 (get the interface configuration for the interface em0) | (the pipe or bar character – next to the letter Z) grep inet (grep is a unix command to search for strings – in this case, the string inet which identifies the IP address) `; (backtick again and then a semicolon to say stop running that command and start doing the next thing) print `date`; (print the response from the command “date” – which returns a date string – and then do the next command) sleep 5 (the sleep command waits integer X number of seconds – here 5, but you could easily say 2 or 10 here – depends on how impatient you and the project manager are!) } (curly brace – meaning to close the for loop from near the beginning) ‘ (apostropy – which instructs it to close the string that the perl interpreter is evaluating)
Notes
[1] from http://stackoverflow.com/questions/555116/repeat-a-unix-command-every-x-seconds-forever – sadly, we can’t use the watch command on a Sidewinder, as it doesn’t have it installed.
[2] Assuming your firewall is in a Primary/Standby configuration (not Standby/Standby), has enabled “Monitor link status” and has addresses to monitor with – presumably the non-HSRP addresses of your attached routers attached to that interface.
