Layer 3 · Go deeper · 7 min

Go deeper

Size it, build it, then make the calls an interviewer will push you on.

Back of the envelope

Size it up

Tap through each step of the arithmetic.

Build-up · 6 steps

One owner beating every second, a three-beat timeout, and a GC pause longer than the timeout. Then what keeps the paused node from doing damage.

Step 1 of 6

Node A owns the work, holds fencing token n and beats every second. The monitor measures silence from the last beat it heard, not from any crash.

4 interview moments

The interviewer pushes on the hard parts. Pick an approach and see how it holds up.

Deep dive 1 of 4
Interviewer
“One node’s GC pauses sometimes outlast your 1 s timeout. It’s declared dead, rejoins and flaps all day. What do you change?”
Your move. Pick an approach:
Deep dive 2 of 4
Interviewer
“You must detect failures across a 5,000-node cluster. Who watches whom?”
Your move. Pick an approach:
Deep dive 3 of 4
Interviewer
“A consumer’s poll loop has hung on one record for minutes, yet the coordinator says it’s healthy. What do you change?”
Its background thread heartbeats every 3 s, and the session timeout is 45 s.
Your move. Pick an approach:
Deep dive 4 of 4
Interviewer
“Your scheduler reassigns a silent worker’s jobs after 30 s. The worker was only paused, and wakes to finish the old job. What stops it?”
Your move. Pick an approach: