Overview

• Kotlin/Ktor implementation of the text-analysis challenge.
• One reusable microservice binary (selected via SERVICE env var) exposes:
  • GET /healthz → { ok: true }
  • POST /op { text?, deps? } → { key, value, cache_hit }
  • GET /debug/pprof/profile?seconds=30 → returns a JFR file (CPU profile)
• Aggregator (POST /analyze { text }) orchestrates calls and returns all fields plus degraded.

Modules

• common: Shared models, text ops, pprof helper (JFR-based)
• service: Reusable microservice. Select behavior with SERVICE env var.
• aggregator: Orchestrates service calls, handles degraded semantics.

Supported Keys

• normalized, transliterated, slug, tokens (array)
• unique_words, bigram_count, char_count, unique_chars
• hash64 (xxhash64), entropy, palindrome

Run (example ports)

• Normalizer: PORT=7001 SERVICE=normalized ./gradlew :service:run
• Transliterator: PORT=7002 SERVICE=transliterated ./gradlew :service:run
• Slug: PORT=7003 SERVICE=slug ./gradlew :service:run
• Tokens: PORT=7004 SERVICE=tokens ./gradlew :service:run
• Unique words: PORT=7005 SERVICE=unique_words ./gradlew :service:run
• Bigram count: PORT=7006 SERVICE=bigram_count ./gradlew :service:run
• Char count: PORT=7007 SERVICE=char_count ./gradlew :service:run
• Unique chars: PORT=7008 SERVICE=unique_chars ./gradlew :service:run
• Hash64: PORT=7009 SERVICE=hash64 ./gradlew :service:run
• Entropy: PORT=7010 SERVICE=entropy ./gradlew :service:run
• Palindrome: PORT=7011 SERVICE=palindrome ./gradlew :service:run
• Aggregator: PORT=7100 ./gradlew :aggregator:run

The Aggregator default service URLs (env override-able): NORMALIZER_URL, TRANSLITERATOR_URL, SLUG_URL, TOKENS_URL, UNIQUE_WORDS_URL, BIGRAM_COUNT_URL, CHAR_COUNT_URL, UNIQUE_CHARS_URL, HASH64_URL, ENTROPY_URL, PALINDROME_URL. Defaults to http://127.0.0.1:<port>/op using the example ports above (7001…7011).

API Examples

• Analyze: curl -s localhost:7100/analyze -H 'content-type: application/json' -d '{"text":"Ångström & co."}' | jq.
• Service call: curl -s localhost:7001/op -H 'content-type: application/json' -d '{"text":"Ångström"}'.
• Health: curl -s localhost:7001/healthz.
• CPU profile (JFR): curl -sLo profile.jfr 'http://localhost:7001/debug/pprof/profile?seconds=15'.

Notes on pprof

• /debug/pprof/profile returns a JFR recording, not Go’s pprof proto. Many tools (JDK Mission Control, async-profiler, Parca via jfr2pprof) can use JFR directly or convert it. If you need native pprof output, we can integrate async-profiler and serve pprof; say the word.

Docker Images (ttl.sh)

• Build & push all service variants plus aggregator (default TTL 24h): TTL=12h PREFIX=myteam bash scripts/publish-ttl.sh
• Each service image sets SERVICE during build; override at runtime if needed: docker run -e SERVICE=hash64 -e PORT=8080 <tag>
• Example aggregator run: docker run -e NORMALIZER_URL=http://normalizer:8080/op -p 8081:8080 <aggregator-tag>
• Generated tags print at the end of the publish script; ttl.sh expires them automatically after the chosen TTL.

---

Multi-arch builds (fix for exec format error)

Some clusters run on ARM64 nodes. If you build an AMD64-only image locally and deploy it to an ARM64 node, the container will fail with:

exec /opt/aggregator/bin/aggregator: exec format error

This repo now builds multi-architecture images (linux/amd64 and linux/arm64) using Docker Buildx.

Prerequisites:

• Docker with Buildx enabled (Docker Desktop or manual):
  • docker buildx version
  • If needed: docker buildx create --use --name multi && docker buildx inspect --bootstrap

Build & push all services + aggregator to ttl.sh as multi-arch:

TTL=24h PREFIX=myteam PLATFORMS=linux/amd64,linux/arm64 \
  bash scripts/publish-ttl.sh

This writes images.env and you can deploy with scripts/deploy.sh.

Quick path: rebuild and redeploy ONLY the aggregator as multi-arch, keeping the current Kubernetes objects:

KCFG=/Users/sondre/repos/cloudnativebergen/2025/blue-independents.kubeconfig
NS=blue-independents
TTL=24h PREFIX=blue-independents \
  bash scripts/deploy-aggregator.sh "$KCFG" "$NS"

The script will:

• Build and push a fresh multi-arch tag to ttl.sh
• Patch the Deployment image and wait for rollout
• Print the Service status

Verification:

kubectl --kubeconfig="$KCFG" -n "$NS" get deploy,po,svc
kubectl --kubeconfig="$KCFG" -n "$NS" logs deploy/aggregator --tail=50

You should no longer see the exec format error. If microservices aren’t deployed yet, /analyze may report degraded=true which is expected until they are running.