The Kid came by my office the other day, positively lit up with excitement about a new log collection tool he just found. If he had a tail it would have been wagging as he asked me to pull up the interface in a browser. The team was already familiar with the product; in fact, it had several features that give pragmatic engineers pause and we'd already rejected it.
It's a tough moment for an IT manager when you're developing talent, but you know you'll have to be a buzzkill and look, again, like an out-of-touch grouch. It was a reminder that we're in an odd, middle-aged epoch in IT. How do you find a sweet spot between adopting evermore tantalizing fads and being so conservative you miss transformative technology? How do we as admins know what's really essential?
"The Kid," Dex, is in his first year on the network team and has earned that affectionate nickname. He just moved up from the helpdesk where he'd been an enthusiastic jack-of-all-trades, closed difficult tickets on the weekends and still had time for R&D. He's an asset to the team and everyone likes him. What he hasn't done however, is build a system from scratch and keep it running, and funded, for a few years. He's a wiz at customizing views and reports in the network monitoring system, but he inherited it from an engineer with years more experience, who in turn inherited it from the senior architect who deployed it in the first place.
Dex did his best to unwind the extra credit customization, like black magic PowerShell and Bash alert automation scripts of his forbears. But he got lost trying to extrapolate the undocumented context behind so much specific, conditional code. While in the midst of creating yet another and particularly unsexy hardware refresh budget spreadsheet, I was thinking about how to help. My first thought was to just tell him to work harder to figure it out. Without guidance, however, it was unlikely he'd know where to start, especially with him gazing so reverently at the Code of Elders. So, at lunch we talked about how to isolate essential requirements from a collage of historical, epic customizations.
I only gave Dex a few suggestions: First, when you inherit a system from someone else, appreciate their investment but realize if it breaks, you will be responsible for fixing it -- even features you don't understand. Second, realize that the former owner was probably as overwhelmed as everyone else on the team and may not have had time to really read release notes on upgrades. Chances are the platform has evolved considerably and you're not using all its available features. Lastly, I recommended he find a quiet whiteboard and while remembering his time on the helpdesk, list out what monitoring was helpful versus what was distracting.
Dex then took a step toward greatness -- he came back with a great plan. Not just great, it was considered and practical. He wanted to do a clean install of the network monitoring solution in the lab and approach it as if he'd never seen it before. He wanted to apply "essentialness" scores to the monitoring inventory and topology, the teams receiving alerts and the backlog of new monitoring requests. From that, he'd grok the platform features and either keep, extend or unwind anything that fell below a certain threshold.
Within about a week, and while juggling his real job of keeping the network lights on, Dex upgraded the way we managed and monitored network and applications. He was able to get the team to reluctantly admit our desktop LAN/data center bias didn't make sense when 80% of the application traffic was either delivered to clients on the WLAN, connected to SaaS/cloud services, or both.
He discovered an unused tool to build access point coverage heat maps that eliminated the weekly mobile audit with a laptop. He made a new friend on the firewall team by consolidating monitoring ACLs for web servers in the DMZ; they'd never been fans of WMI. He even helped me by creating an evidence-based capacity forecast report that I could roll up with my refresh plan. Best of all, Dex did it with $0 in new spend.
The Kid stays in the picture
It's sometimes easy for IT veterans to roll our eyes should a lumbernetworkjack hipster admin roll in with an awesome but unproven approach. On the other hand, just because we might remember Token Ring hermaphroditic IBM data connectors and the eccentricities of Crescendo XDI before it became CatOS does not mean that what's tried and true today will support our IT mission five years from now. But if you focus on what's truly essential to monitor, you'll naturally balance new hotness with old and reliable.
But enough about that. Did you see this new NoSQL relational mapping engine? It's made from artisanal, recycled electrons, but you've probably never heard of it.