Datadog · Primly Community

Went through the Datadog SRE loop last month, here's what actually matters

infra_ines · 5 replies

Finished my Datadog SRE loop about 5 weeks ago. Sharing because I found basically nothing useful when I was prepping.

The loop was 5 rounds virtual: coding screen, then four back-to-back sessions on a single day. Sessions were: a coding round (medium leetcode, nothing crazy), a system design round, a "debugging" round, and a behavioral. There was also a short intro call with the hiring manager.

The debugging round is the one nobody talks about. They gave me a simulated scenario: a service with degraded throughput, and I had access to what was supposed to look like a simplified Datadog UI with metrics, logs, and traces. I had to walk through what I'd check and why. Not memorizable. You have to actually know how you investigate production issues. I kept narrating my reasoning and I think that's what saved me. Silence is bad in that round.

System design was infra-flavored. Not "design Twitter." More like "how would you build a metrics ingestion pipeline that handles spikes without dropping data." Really specific to what Datadog actually does. Expect tradeoffs around cardinality and write throughput.

Behavioral was low-key but they probed on incident response stories specifically. "Tell me about a time a system you owned had an unexpected failure" came up. Have a real story, not a polished one.

One thing that caught me off guard: the interviewers were really senior. Like, principal+ level engineers in the room. They can tell when you're pattern-matching to a rehearsed answer. Just be direct.

5 replies

sre_sol

the debugging round sounds exactly like the kind of thing i'd either ace or completely brick depending on sleep quality. was it open-ended or did they tell you what the problem was?

infra_ines

open-ended, they just said "this service is slow, investigate." you had to figure out what slow even meant and where to look. i started with the latency histogram and worked outward. no hand-holding.

remote_swe_42

did you negotiate? curious how rigid they are on initial offers for SRE roles.

infra_ines

yes, and they moved. not dramatically, but they had room in the RSU component. total comp was the lever, not base.

backend_bekah

the metrics-ingestion design question is classic Datadog. they basically want you to design a simplified version of what they built. useful to know they use kafka-style pipelines under the hood and care a lot about backpressure handling.