Clock Skew

ELI5 - The Vibe Check

Every server has its own clock, and those clocks drift apart from each other over time, sometimes by milliseconds, sometimes by embarrassingly more if NTP sync breaks. Clock skew is that drift, and it's a genuinely dangerous problem in distributed systems, because a lot of logic quietly assumes 'this timestamp happened before that one' is a fact you can trust, right up until two servers disagree about what time it even is.

Real Talk

Clock skew is the difference in time between clocks on different machines in a distributed system, arising from imperfect NTP synchronization, hardware drift, or virtualization overhead. It undermines any logic that relies on wall-clock timestamps for ordering events across nodes, including lease expiry, cache TTLs, and log correlation, which is why systems requiring strict event ordering use logical clocks (Lamport timestamps, vector clocks) or hybrid logical clocks instead of trusting wall-clock time alone.

When You'll Hear This

"The distributed trace looked broken because clock skew made a child span appear to start before its parent." / "We stopped trusting wall-clock ordering entirely and switched to Lamport timestamps."