Slashdot Mirror
System Admins Should Know How To Code
snydeq writes "You don't need to be a programmer, but you'll solve harder problems faster if you can write your own code, writes Paul Venezia. 'The fact is, while we may know several programming languages to varying degrees, most IT ninjas aren't developers, per se. I've put in weeks and months of work on various large coding projects, but that's certainly not how I spend most of my time. Frankly, I don't think I could just write code day in and day out, but when I need to develop a tool to deal with a random problem, I dive right in. ... It's not a vocation, and it's not a clear focus of the job, but it's a substantial weapon when tackling many problems. I'm fairly certain that if all I did was write Perl, I'd go insane.'"
PS: YOu don't really need to know how to program, if you can just identify tasks that are braindead and repetitive, that's already a plus. Then you can go and talk to someone who does know how to program, explain the problem, and this person might come up with a simple solution for you. It all boils down to this.
Unfortunately, many people are not trained to identify automatable jobs.
"It's too bad that stupidity isn't painful." - Anton LaVey
You don't or you can't? Because even when I'm not coding, just in day-to-day computer use, I often write scripts to batch-process or automate various things. I have found this highly beneficial. I don't want to break you down, I just want to know your reasoning. Personally, I can highly recommend learning scripting, be it bash/awk/php (yes, I wrote php scripts)/powershell/whatever, I'm sure you'll benefit from it.
"It's too bad that stupidity isn't painful." - Anton LaVey
Exactly - it comes down to a matter of cost very often. If you can spend a half hour throwing together a script to make a configuration change across, say, 2000 devices (or in our case, several million), it is far cheaper than trying to find a vendor solution, or having folks go out and do the work themselves. The vendor will often have maintenance fees and high initial costs for a tool that "sort of" does what you want. You can have people go out and pound pavement, but if you have 2000 devices and send out 50 people, and it takes each person 20 minutes to do the work on the device plus 30 minutes trave time, times 17$ an hour... well, that adds up fast too. Not to mention the opportunity cost and backlog of tickets that that would generate.
To be honest, most of my managers have been very nervous about my suggestions to even use tools which I'm knowledgable in because they don't understand the concepts. Ad the bigger the company the more likely they are to want a vendor solution rather than use internal resources.
As a manager, sometimes I see the tool to automate and script changes to be a risk of taking down the network faster than anyone can do anything about it.
Like the time a senior network admin accidentally took down our network by setting the IP address of all of our network switches to the same set of IP address. He tested the script and it worked perfectly on one network closet - then he proceeded to apply the remaining set of IP addresses the the remaining hundred network switches in our organization.
He knew as soon as the first alarm went off what had happened, but by then it was too late to stop it. Fortunately, he had the good sense to not write the config to memory, so recovery just walking to each network closet to power cycle the switches -- much faster than if we had to walk around with a console cable to back out the changes on each switch
That said, he still uses the same tool to push out changes, but he tests his scripts on 2 IDF's before pushing out changes and has the junior network admin double check his work.
But I can see why a manager would be cautious about a new and untested automation tool.