Fara , ba daga ƙasa ba daga aiki-aiki. Jerin masipati mafi aminci shine bakin ciki idan bukatun suke kasa, saboda haka fara da da aka gabatar ga masu amfani — , , — sannan ƙari da sigina mai kyau hudu, sannan sigina aiki-aiki ta ƙarshe.
Fara , ba daga ƙasa ba daga aiki-aiki. Jerin masipati mafi aminci shine bakin ciki idan bukatun suke kasa, saboda haka fara da da aka gabatar ga masu amfani — , , — sannan ƙari da sigina mai kyau hudu, sannan sigina aiki-aiki ta ƙarshe.
1. USER-FACING SLIs → what the user experiences (latency, errors, availability)
2. GOLDEN SIGNALS → latency, traffic, errors, saturation per service
3. INFRA METRICS → CPU, memory, disk, network (causes, not symptoms)
Ko da kun kalba CPU da diski kawai (daga ƙasa), zaku iya zama sananne gida yayin da masu amfani ke samun 500s. Kallonka SLI da farko (daga saman) yana nufin za ku yi offaa da alamar masu amfani suke ji ainihi, sannan ku nutsar zuwa sigina mai kyau da aiki-aiki don nemo dalili.
INSTRUMENT app emits metrics/logs/traces (e.g. request_duration_seconds histogram)
↓
COLLECT a TSDB scrapes/ingests them (Prometheus, Datadog agent)
↓
DASHBOARD visualize SLIs + golden signals (Grafana) for humans to read
↓
ALERT fire on SLO violations / burn rate, routed to on-call
# Availability SLI: fraction of requests that succeed
sum(rate(http_requests_total{status!~"5.."}[5m]))
/ sum(rate(http_requests_total[5m]))
# Latency SLI: p99 request latency
histogram_quantile(0.99, sum by (le) (rate(http_request_duration_seconds_bucket[5m])))
Sana'a SLO akan kowane SLI (misali 99.9% kasancewa, p99 < 300ms), allon gida su, da offaa lokacin da SLO yana bakin haure — ba akan kowane waje ba.
Sani-sani da aka gina daga ƙasa tana gaya maka diski shine 80% cikakke amma ba gaya ba cewa masu kasuwa ba za su iya kammala saye ba. Fara daga SLI da aka gabatar ga masu amfani yana danganta kila allon gida da offaa zuwa tasiri ainihi na masu amfani, yana kiyaye jiya ƙasa, da yana ba da hanyar fili (alamar → sigina mai kyau → dalili aiki-aiki) lokacin da wani abu ya karye.