What should I actually check before assuming a slow API response is my app's fault?

Question

My API on a DigitalOcean Droplet has been responding slowly for some requests, sometimes 2-3 seconds when it should be under 200ms. I keep assuming it is my code, but before I spend hours profiling everything, is there a sensible order to rule out network, DNS, database, or Droplet-level issues first?

Bobby · Answer

Hi there, Good instinct to check the layers in order rather than jumping straight to code profiling. Here is a sensible order, cheapest checks first. First, rule out DNS resolution delay, this is an easy one to miss and surprisingly common: dig yourapi.com Look at the query time at the bottom of the output. If DNS itself is slow, that adds latency before your app even gets a request. Next, check raw network timing with curl, this separates connection time from your app’s actual processing time: curl -w "dns=%{time_namelookup} connect=%{time_connect} ttfb=%{time_starttransfer} total=%{time_total}
" -o /dev/null -s https://yourapi.com/endpoint If ttfb (time to first byte) is high but connect is fast, the delay is happening inside your app or database, not the network. If connect itself is slow, that points at Droplet-level network issues or a firewall doing extra work. Check if the slowness is consistent or intermittent: for i in {1..10}; do curl -w "%{time_total}
" -o /dev/null -s https://yourapi.com/endpoint; done Consistent slowness points at a real bottleneck somewhere. Intermittent slowness often points at resource contention, database connection pool exhaustion, or another process competing for CPU on the same Droplet. Check your database next, since a slow query is the most common cause of API latency: -- Postgres SELECT pid, now() - query_start AS duration, query FROM pg_stat_activity WHERE state = 'active' ORDER BY duration DESC; Anything running longer than your target response time is your culprit. Check Droplet resource pressure itself: top free -h CPU maxed out or memory swapping will slow down everything indiscriminately, and this can look like “my app is slow” when it is actually “my Droplet is undersized for the load.” Only after ruling out DNS, network, database, and Droplet resources should you start profiling your application code directly. Working top-down like this saves you from spending hours optimizing code when the real issue is a missing database index or a Droplet that needs upgrading.

alexdo · Answer

Heya, Bobby’s order is right, two refinements. That curl format string is missing the TLS handshake, which matters since your endpoint is HTTPS. Without it, handshake time gets absorbed into what looks like app time: curl -w “dns=%{time_namelookup} connect=%{time_connect} tls=%{time_appconnect} ttfb=%{time_starttransfer} total=%{time_total}
” -o /dev/null -s https://yourapi.com/endpoint Read it as differences: tls minus connect is the handshake, ttfb minus tls is your app actually thinking. That last number is the one to judge before blaming your code. Second, in the Droplet resource step, check steal time, not just CPU and memory. Run vmstat 1 5 and look at the st column. Steal is CPU your Droplet wanted but did not get because the host gave it elsewhere, and top’s load average will not show it. This matters if you are DigitalOcean’s plan docs saythe hyper-thread “may be shared between multiple other Droplets” and that underneighbour load “a Droplet coyper-threads.” Their ownadvice there is to move to dedicated CPU if variable performance is intolerable. Non-zero steal plus low CPU of your own, with intermittent spikes, is a recognisable shape, and it i it is not your code. Regards

Become a contributor for community

DigitalOcean Documentation

Resources for startups and AI-native businesses

The developer cloud

Start building today

Report this

What should I actually check before assuming a slow API response is my app's fault?

Related questions

I followed the installation process and my sites doesn't work!

App R Shiny time-out

Digital Ocean Infrastructure

error accessing external endpoint kubernetes in digital ocean

Does droplet support SCTP protocol?

Uncaught ReferenceError: require is not defined

Start Qt Designer application? How? Where?

[CRITICAL] WORKER TIMEOUT - How can i reach gunicorn timeout settings?

Snowflake duplicates

how to expand memory

How can I know my price of hosting web app

Which product from Digital Ocean is better for hosting big applications?

Report this

What should I actually check before assuming a slow API response is my app's fault?

Related questions

I followed the installation process and my sites doesn't work!

App R Shiny time-out

Digital Ocean Infrastructure

error accessing external endpoint kubernetes in digital ocean

Does droplet support SCTP protocol?

Uncaught ReferenceError: require is not defined

Start Qt Designer application? How? Where?

[CRITICAL] WORKER TIMEOUT - How can i reach gunicorn timeout settings?

Snowflake duplicates

how to expand memory

How can I know my price of hosting web app

Which product from Digital Ocean is better for hosting big applications?