[technofiend]

It saves so many headaches. I used to support a system with an automated fallback. If today's feed wasn't in by cutoff then the previous day's feed was used.

Unfortunately we'd sometimes get partial or corrupted feeds. Partial feeds triggered investigation and possibly a manual rerun and corrupted ones often halted the system.

Because we only used monthly numbers for reporting, delaying and rerunning any other day was pointless beyond standard root cause analysis to prevent recurrence. And this system had hundreds of feeds so at first there were almost daily issues.

So I added a check to throw out any deviations over two sigma from the median of the last 30 days' good feeds which knocked out 99% of our data quality issues. I got in a boatload of trouble for different reasons but that's another story.

[ThePowerOfFuet]

>I got in a boatload of trouble for other reasons but that's another story.

You cannot just drop that line and walk away. Storytime?

[technofiend]

This predated DevOps: we were using the Carnegie Mellon Capability and Maturity Model (CMM) which split roles into Plan, Build, Operate and Control with me in the Operate tier. I proposed the change, got Build to schedule and release it but I didn't get buy in from the architect of the system. He was toweringly, incandescently angry that I had dared touch his design. To the point where I think he tried to get me fired over it. Fortunately cooler heads prevailed because really I was right and it was all just an ego thing on his part. I took a page from my boss who was masterful at dealing with these kinds of things. We buried the architect in all things operational to the point where he eventually asked us to stop and grudgingly allowed we knew what we were doing. I don't know if he believed it or just wanted the emails to stop.