English for Tech
Amélie

English for DevOps: incidents, postmortems, runbooks

Production breaks at 3am in a multi-timezone team. Your accent barely matters—but missing the stress patterns in 'We've got a SEV1' means you're seconds behind the conversation. Real DevOps English is faster, blunter, and structurally different from textbook English.

Try Amélie free →

Why this matters

Most French tech professionals know the word 'incident' but freeze when an engineer says 'We're looking at a cascade failure in auth—redundancy held for 8 minutes but now checkout's erroring.' The gap isn't vocabulary; it's the stress patterns, passive-voice strategies, and structural choices that multinational DevOps teams use under pressure. Postmortems demand different language—blameless retrospectives require specific tense, prepositions, and causality markers that French L1 speakers often misalign. Add runbooks (imperative, clipped, few articles) and incident-call urgency (direct, heavily stressed), and you're navigating three registers of English at once. This course teaches the exact phrases, stress patterns, and structures real engineers use from first alert to postmortem.

You're 90 minutes into a production incident affecting EU users. Your Slack channel is moving at 200 messages/minute. Someone posts: 'We're seeing a cascade failure triggered by the database connection pool timeout—rolling back the 14:15 deployment.' You understand each word, but the structure of the sentence ('seeing a cascade failure triggered by...') doesn't click immediately. By the time you've parsed it, the team's already moved to 'What's our rollback ETA?' and you're lost.

Practical tips

Master stress patterns in incident calls

Incident communication isn't natural English—certain words get heavy stress to signal urgency. 'We've got a SEV2' sounds different from 'We've got a sev-two.' Native engineers stress SEV levels, acronyms, and numbers differently than French L1 learners expect. Train your ear on real incident recordings and mirror the rhythm.

Passive voice dominates postmortems

Runbooks use imperative ('Restart the service'), incident calls use direct present ('We're seeing 50% errors'), but postmortems use passive past ('The service was restarted without rolling back the database migration'). This passive structure removes blame and signals a blameless culture. French postmortems often sound active and accusatory—learn the passive pattern or you'll sound like you're blaming someone.

Know severity levels and when they're used

SEV1 (all hands on deck, CEO gets paged), SEV2 (major impact, team mobilizes), SEV3 (minor, on-call handles it). Each triggers different language and urgency. If you respond to a SEV1 with the tone of a SEV3, you'll be ignored. Understand the unwritten rules: SEV1 uses imperative + present tense, SEV3 uses softer language and future planning.

Learn rapid-fire technical structures

Incident engineers use compressed syntax: 'We're in a state where the database is timing out' instead of 'The database experiences timeouts.' Or: 'Traffic to the EU region is degrading' vs 'Users in the EU experience slower responses.' These structures pack diagnosis into the noun phrase itself. Practise building these structures—they're non-negotiable in fast conversations.

Root cause is always singular in English

French speakers often say 'causes' because multiple factors contributed. But postmortems isolate one root cause ('Insufficient rate limiting') and list contributing factors separately ('High traffic,' 'New code deployment'). This distinction matters for blameless culture—you're separating 'why the system failed' from 'what conditions enabled it.'

Runbooks are telegraphic—minimal grammar

A runbook isn't prose. It's: 'Check metrics on dashboard. If CPU > 80%, scale up. Wait 2 minutes. Verify health endpoint responds.' No articles, no 'the,' no extra words. French learners often add too much structure. Strip it down. Every word must be actionable.

Incident calls and postmortems use opposite registers

Incident calls are command-and-control: 'Get the on-call DBA. Check the logs. Report in 60 seconds.' Postmortems are collaborative and explanatory: 'The team identified that rate limiting was insufficient. We deployed a fix that reduced retry storms by 85%.' Same event, completely different English. Practise both.

Master cascade, redundancy, and graceful degradation

These three words unlock DevOps conversations. 'Cascade failure' (one thing breaks, then another, then another), 'redundancy held the impact' (backup systems worked), 'graceful degradation' (you got reduced functionality instead of total loss). Native engineers use these constantly. Understand them cold.

Phrases natives use

Opening an incident call
We're actively monitoring the situation. What we know so far: the auth service is returning 500s to 15% of requests.
French speakers often use 'We are monitoring'—omit the continuous aspect to sound native.
Asking for help during an incident
Can you get eyes on the database replication lag? We're seeing delays on the replica.
'Get eyes on' is idiomatic for 'look closely at'—common in English teams, rare in French-taught English.
In a postmortem
The root cause was a deployment that lacked sufficient rate limiting on the endpoint.
Passive + past tense + nominalised verbs ('a deployment that lacked') is the postmortem register. French tends toward active.
Explaining impact in a postmortem
Although redundancy held for the first 8 minutes, the circuit breaker didn't trip fast enough, so we saw a cascade failure.
The 'although...so' structure signals contributing factors + outcome—essential for blameless postmortems.
Runbook instruction
Restart the auth service. Wait 30 seconds. Curl the health endpoint. If status is 200, proceed. Otherwise, escalate to on-call DBA.
Imperative, present tense, minimal articles—telegraphic style. French learners over-explain.
During an incident call
We're seeing a cascade failure triggered by the connection pool exhaustion. Rollback is our best option here.
'Triggered by' is more idiomatic than 'caused by'—native engineers prefer it in fast speech.
In a status update
We've implemented better observability around this component. Future incidents will be faster to diagnose.
Perfect present (We've implemented) + future (will be) is standard for action items. French learners often use simple past.
Explaining mitigation
The service gracefully degraded to read-only mode. Users saw slower responses but not complete loss of functionality.
Graceful degradation is nearly untranslatable from French—native concept specific to English DevOps culture.
Severity marker
This is a SEV2. All engineers on this service need to be involved.
SEV levels are spoken as letters ('S-E-V-two') and carry immediate cultural weight. French learners often miss the urgency.
Incident wrap-up
We're standing down from the incident. The on-call engineer will monitor for the next two hours and escalate if anything recurs.
'Standing down' (military/emergency term) is the standard phrase for ending an incident—not 'closing' or 'finishing.'

FAQ

What's the difference between 'incident,' 'outage,' and 'issue'?

An 'issue' is any problem (code bug, slow response). An 'outage' is complete loss of service (EU region is down, can't log in). An 'incident' is the operational response to either—the event, the team mobilization, the postmortem. You'll hear 'We have an incident' not 'We have an outage'—incident refers to the whole response, not just the downtime.

Why do postmortems use so much passive voice?

Passive voice removes the actor, which signals a blameless culture. 'The service was restarted' (passive, no blame) vs 'John restarted the service' (active, implies John made a mistake). English DevOps teams use passive to separate the action from the person—postmortems focus on systems and processes, not individuals. This is nearly absent in French business writing.

How do I sound confident during incident calls when English isn't my first language?

Use simple, short sentences. 'Database is slow. I'm checking metrics. I see CPU at 92%.' This is 100% natural in incident calls—speed matters more than grammar. Native engineers code-switch to this clipped style automatically. Don't try to sound fluent; sound functional. And use 'I' statements for what you're doing: 'I'm pulling the logs' (not 'the logs are being pulled').

What vocabulary should I prioritize for DevOps English?

Severity levels (SEV1, SEV2, SEV3), incident states (ongoing, mitigated, resolved, postmortem), technical structures (cascade failure, redundancy, graceful degradation, circuit breaker, rate limiting), tenses (present continuous for what's happening now, perfect for what's been done, passive past for postmortems), and register shifts (imperative for runbooks, passive for postmortems, direct for incident calls). Prioritize in that order.

Amélie understands your L1

The only AI English coach that catches L1 transfer errors. 19,99€/mo — first session free.

Get started →