Workflow Error Tracking: Sentry vs Bugsnag vs Rollbar

TL;DR

• Sentry: ekosistem terbesar, performance + replay + profiling all-in-one. Default untuk SaaS multi-channel.
• Bugsnag: mobile-first, stability score matang. Conditional untuk mobile-heavy product.
• Rollbar: workflow team-led triage, deploy tracking solid. Conditional untuk team yang process-driven.
• Datadog APM (alternative): bundle dengan observability tapi cost premium.
• Verdict: Conditional — Sentry default untuk Indonesia mid-market, self-host kalau compliance ketat.

Konteks

Saya jalankan ketiga platform paralel 22 bulan (Agustus 2024 - Mei 2026) di fintech series-B Jakarta. Profile:

• 18 microservice backend (Java, Go, Node, Python)
• 6 frontend channel (web React, web admin, mobile Android Kotlin, mobile iOS Swift, mobile Flutter cross, BI dashboard)
• 18.500 MAU total
• ~250k error event/bulan (campuran exception sungguhan + warning yang logged)

Setup paralel saya:

• Sentry self-host di GKE Jakarta untuk backend + web + admin (production primary)
• Bugsnag SaaS untuk mobile Android/iOS native (heritage stability score)
• Rollbar SaaS eval 6 bulan untuk satu domain service, kembalikan ke Sentry karena consolidate vendor

Sebelumnya di BUMN 2022-2024 saya pakai Rollbar 14 bulan. Pengalaman itu masuk pembanding.

Pricing (Juni 2026)

Sentry

• Developer (Free): 5k error event/bulan, 10k transaction (performance), 50 replay
• Team: USD 26/bulan + USD 0,29/1k event extra
• Business: USD 80/bulan + same overage
• Enterprise: quoted, mulai USD 2.000/bulan untuk seat banyak + SSO/SAML
• Self-host (open source): gratis software (BSL license dengan free for production use), infra cost

Saya pakai self-host. Cost infra:

• GKE 3 pod e2-medium: Rp 2,1 juta/bulan
• Postgres db-g1-small: Rp 720 ribu/bulan
• Redis basic 1GB: Rp 350 ribu/bulan
• ClickHouse self-host 1 VM e2-medium (untuk performance data): Rp 720 ribu/bulan
• Storage symbol + artifact di R2: Rp 200 ribu/bulan
• Total Sentry self-host: Rp 4,1 juta/bulan + ops 10-15 jam/bulan

Bugsnag

• Free: 7.5k event/bulan, 1 project
• Lite: USD 59/bulan, 100k event
• Standard: USD 199/bulan, 1M event
• Enterprise: quoted
• Saya pakai Standard untuk 2 mobile native (Android + iOS): USD 199/bulan = Rp 3,2 juta/bulan

Rollbar

• Free: 5k occurrence/bulan
• Essentials: USD 21/bulan, 50k
• Advanced: USD 79/bulan, 200k
• Enterprise: USD 499/bulan, 1,5M + SSO
• Saya eval 6 bulan di Advanced tier: USD 79/bulan = Rp 1,3 juta/bulan

Total bundle aktif

Platform	Tier	Cost/bulan
Sentry self-host	open source	Rp 4,1 juta
Bugsnag SaaS	Standard	Rp 3,2 juta
Total		Rp 7,3 juta/bulan

Bandingkan kalau full Sentry SaaS Business: USD 80 base + USD 50 overage 250k event + USD 30 replay sample = USD 160/bulan = Rp 2,56 juta/bulan. Lebih murah dari self-host kalau cost cash flow saja yang dilihat.

Tapi self-host saya advantage: data residency Indonesia + tidak ada per-event charge yang scale. Untuk fintech regulated, worth premium.

SLO + performance

Metrik	Target	Realisasi Sentry self-host	Bugsnag SaaS
Ingest availability	> 99,9%	99,93%	99,98%
Event-to-dashboard latency p99	< 30 detik	18 detik	12 detik
Alert latency event-to-Slack	< 60 detik	42 detik	35 detik
Symbolication mobile crash	< 5 menit	3,5 menit	2,8 menit
Source map upload success	> 99,5%	99,8%	99,7%

Insiden 22 bulan:

• Sentry self-host 4 P2 incident (Postgres lock, ClickHouse memory spike, deploy regression)
• Bugsnag 0 incident
• Rollbar 1 incident di periode eval (vendor outage 28 menit)

Capability comparison

Capability	Sentry	Bugsnag	Rollbar
Backend error tracking	ya	ya	ya
Frontend JS error tracking	ya	ya	ya
Mobile native (iOS/Android)	ya	ya (heritage)	ya
React Native / Flutter	ya	ya	ya
Performance monitoring (APM)	ya (built-in)	terbatas	tidak
Session replay	ya	tidak	tidak
Profiling	ya	tidak	tidak
Stability score (mobile)	ya	ya (best)	terbatas
Deploy tracking	ya	ya	ya (workflow-strong)
Assignment + workflow	ya	ya	ya (best)
Source map / symbol upload	ya	ya	ya
Self-host option	ya	tidak	tidak
Data residency Indonesia	self-host yes	tidak	tidak
Audit log compliance	enterprise tier	enterprise tier	enterprise tier
Pricing predictability	medium (per-event)	medium (per-event)	medium (per-occurrence)
Free tier dermawan	sedang	sedang	sedang
Integration SDK count	100+	50+	60+

Mobile crash reporting

Bugsnag heritage di mobile crash. Pengalaman saya 22 bulan:

Aspek	Sentry mobile	Bugsnag mobile
Stack trace iOS Swift	bagus	excellent (lebih detail context)
Stack trace Android Kotlin	bagus	excellent
Native crash (C/C++ via JNI)	OK	excellent
Symbolication speed	3,5 menit	2,8 menit
Stability score metric	sederhana	matang (per release, per user segment)
Release health dashboard	bagus	bagus
ANR detection (Android)	ya	ya

Untuk mobile-heavy product, Bugsnag stability score lebih useful untuk product team. Untuk mobile companion (frontend secondary), Sentry cukup.

Backend + frontend web

Sentry menang untuk backend + web SPA. Coverage SDK luas, performance tracing native, session replay unique. Bugsnag backend SDK ada tapi feature lebih sederhana.

Trade-off arsitektural

Pilih Sentry kalau:

• Multi-channel (backend + frontend + mobile) — coverage unified
• Butuh performance monitoring + replay + profiling all-in-one
• Compliance ketat — self-host option penting
• Tim engineering > 10 — feature breadth Sentry justify learning curve
• Cost-sensitive di mid-scale (Self-host Sentry vs Datadog: 10x lebih murah)

Pilih Bugsnag kalau:

• Mobile-heavy product (50%+ traffic mobile native)
• Stability score per release adalah KPI utama
• Workflow product team (release manager pakai data crash untuk go/no-go)
• Budget premium OK

Pilih Rollbar kalau:

• Workflow team-led triage prioritas (assignment, snooze, RCA template)
• Deploy tracking sebagai first-class citizen
• Tim kecil-medium (1-15 dev) tanpa kebutuhan APM/replay

Pattern multi-vendor

Bundle saya (Sentry self-host + Bugsnag SaaS mobile): trade-off cost vs depth per-channel. 2 vendor = 2 tooling chain. Worth-nya hanya kalau mobile native dominan (60%+ traffic) — kalau tidak, Sentry single-vendor lebih sederhana.

Sampling strategy

Untuk control cost + noise:

Sentry sampling

# Backend sampling
sentry_sdk.init(
    sample_rate=1.0,  # 100% error
    traces_sample_rate=0.1,  # 10% performance trace
    profiles_sample_rate=0.1,  # 10% profile
    replays_sample_rate=0.05,  # 5% session replay (web)
    replays_on_error_sample_rate=1.0,  # 100% replay saat error
)

Pattern: error 100% (kita butuh semua), performance 10% (statistik cukup), replay 5% baseline + 100% saat error.

Filter noise

// Frontend
Sentry.init({
  ignoreErrors: [
    'Network request failed',  // mobile offline noise
    'NotAllowedError',  // user denied permission
    /Loading chunk \d+ failed/,  // hot reload artifact
  ],
  denyUrls: [
    /chrome-extension/,  // extension noise
    /safari-extension/,
  ],
});

Saya filter ~35% event noise. Total event ter-ingest turun dari ~380k ke ~250k/bulan = save USD 38/bulan (kalau SaaS).

Fingerprint grouping

Agresif group error mirip. Default Sentry group by stack trace top frame — kadang terlalu granular. Custom fingerprint:

Sentry.init({
  beforeSend(event) {
    if (event.exception?.values[0]?.type === 'TimeoutError') {
      event.fingerprint = ['timeout-{{ default }}'];
    }
    return event;
  },
});

Reduce issue count dari ~1.200 unique ke ~340 unique. Tim triage capacity catching up.

Compliance + data residency

Untuk fintech Indonesia

Sensitive data di error event:

• Request payload (potentially PII)
• Stack trace yang berisi local variable value
• User context (email, user ID)
• IP address

Sentry self-host advantage: semua data tinggal di GKE Jakarta + R2 Jakarta. UU PDP 2022 compliant. OJK audit Q1 2026 passed.

Untuk Sentry SaaS: data ke US server. Untuk Bugsnag/Rollbar SaaS: similar. PII scrubbing built-in (regex pattern), tapi tidak 100% reliable. Untuk service yang menyentuh data nasabah: self-host non-negotiable.

PII scrubbing pattern

sentry_sdk.init(
    send_default_pii=False,
    before_send=scrub_pii,
)

def scrub_pii(event, hint):
    if 'request' in event:
        # Mask email, phone, NIK, account number
        event['request']['data'] = mask_sensitive(event['request']['data'])
    return event

Maintain regex pattern Indonesia-specific: NIK 16-digit, NPWP 15-digit, nomor rekening 10-16 digit, nomor HP +62/08 prefix. Custom scrubber 1-2 hari kerja initial.

Workflow + integration

Sentry workflow

• Issue auto-create di Linear / Jira via integration
• Slack notification per-issue (custom rule per severity)
• Deploy tracking via release API + commit hash
• Code mapping (Sentry → GitHub source)

Bugsnag workflow

• Stability score threshold trigger (auto-block release kalau crash rate > 1%)
• Mobile release health dashboard
• Slack + PagerDuty integration

Rollbar workflow

• Assignment + snooze workflow strongest
• Deploy tracking dashboard sederhana clean
• RCA template tertanam di workflow

Untuk team yang process-driven (review error mingguan, assignment formal), Rollbar workflow lebih ergonomis. Untuk team yang feature-led, Sentry cukup.

Cost of ownership 36 bulan

Skenario: SaaS B2B fintech Indonesia, 250k event/bulan, multi-channel.

Item	Sentry self-host + Bugsnag	Sentry SaaS Business + Bugsnag	Datadog APM full
Subscription/infra 3 tahun	Rp 263 juta	Rp 95 juta + Rp 115 juta	Rp 2,1 miliar
Ops time engineer	Rp 130 juta	Rp 24 juta	Rp 18 juta
Migration cost (one-time)	Rp 30 juta	Rp 0	Rp 60 juta
Total 3 tahun	Rp 423 juta	Rp 234 juta	Rp 2,18 miliar

Self-host actually lebih mahal dari SaaS karena ops overhead. Justify-nya bukan cost — justify-nya data residency compliance. Untuk SaaS non-regulated, Sentry SaaS Business + Bugsnag mid-tier = sweet spot.

Common pitfall

1. Event volume explosion. Loop error tanpa rate limit = ribuan event identik per detik. Implement client-side rate limit per fingerprint (max 100 event/menit per error).
2. Source map tidak ter-upload. Tanpa source map, stack trace minified = useless. Setup CI step upload source map ke Sentry / Bugsnag pada setiap deploy. Verify dengan test event di CI.
3. Sensitive data leak via context. Stack trace berisi variable value = potentially PII. Scrub agresif via beforeSend hook + audit periodically.
4. Tidak ada Triage routine. 1.200 unique issue tanpa triage = noise. Setup weekly triage 30 menit + assignment per domain owner.
5. Alert fatigue. Alert per-error = inbox spam. Pakai alert threshold (only alert kalau frequency > 10/menit, atau new issue belum pernah seen) + on-call rotation.

Indonesia specific

Mobile offline noise

Banyak mobile user Indonesia jaringan tidak stable (4G di tier-2 city). Network error volume tinggi. Filter di SDK:

Bugsnag.addOnError { event ->
    val error = event.errors[0]
    if (error.errorClass == "java.net.UnknownHostException") {
        return@addOnError false  // skip
    }
    true
}

Saves 30-50% event volume mobile.

Compliance OJK

OJK Cyber Resilience Guideline mensyaratkan log retention 12 bulan + audit trail. Sentry self-host: configure retention 18 bulan, export ke S3 Glacier untuk 7 tahun archive. Cost retention archive: Rp 600 ribu/bulan ongoing + Rp 8 juta amortized one-time setup.

Yang surprising

Setelah 22 bulan: Sentry session replay ternyata ROI tinggi untuk debug UI bug. Bug yang dulu butuh 2-3 hari investigation + customer support back-and-forth, sekarang 15-30 menit dengan replay. Saya naikkan replay sample rate dari 5% ke 10% untuk web product.

Surprise lain: Bugsnag mobile stability score actually mengubah cara product team release. Sebelum: release setiap 2 minggu independent dari crash rate. Setelah: release blocked kalau stability score di bawah threshold (99,5% untuk Android, 99,7% iOS). Crash rate stable improve dari 1,2% ke 0,4% dalam 8 bulan.

Verdict

Conditional dengan rule konkret:

• Default SaaS multi-channel Indonesia: Sentry. Coverage end-to-end terbaik.
• Self-host Sentry kalau compliance ketat (fintech, healthcare, govt). Cost Rp 4-5 juta/bulan infra.
• Sentry SaaS Business untuk SaaS non-regulated. Cost USD 80-300/bulan.
• Tambah Bugsnag kalau mobile native > 50% traffic + stability score kunci KPI. Worth multi-vendor overhead.
• Skip Rollbar untuk SaaS modern Indonesia kecuali workflow team-led adalah prioritas extreme.

Threshold konkret untuk adopt platform error tracking: pengguna pembayar > 50 atau MAU > 1k. Di bawah ini, raw log ke Loki + grep cukup. Platform overhead tidak justify untuk MVP.