📢 Steemit Platform Stability Update 0406

Issue Summary

Over the past week, Steemit users experienced intermittent instability and slow page loads during the daily window of 02:00–04:00 CST (UTC+8). Our monitoring identified elevated API response times and database query bottlenecks during this period.

---

Root Cause Analysis

Our investigation revealed two primary bottlenecks:

Backend (hivemind)

• Slow database queries for trending, feed, and community APIs
• Unused database indexes consuming ~2GB memory without performance benefit

Frontend (condenser)

• SSR (Server-Side Rendering) making 3–5 sequential API calls with zero caching
• Average SSR latency: 218–340ms (threshold: 150ms)
• specialPosts() function re-fetching 10+ blockchain API calls on every request

---

Fixes Deployed

Backend Optimizations (hivemind)

PR #368 – Query Optimization

• ✅ Added 4 new database indexes for trending/feed/community queries
• ✅ Replaced NOT IN subqueries with NOT EXISTS for better PostgreSQL performance
• Expected improvement:
  • Trending queries: 281ms → ~20ms (14x faster)
  • Feed queries: 30ms → ~5ms (6x faster)
  • Community queries: 23ms → ~3ms (7x faster)
  • Total CPU usage: ~13.6% → ~2%

PR #367 – Index Cleanup

• ✅ Dropped unused hive_posts_cache_ix31 index (~2GB freed)
• ✅ All hot queries now route through optimized temp tables

Frontend Optimizations (condenser)

PR #3974 – API Caching & Parallelization

• ✅ Added ApiCache with TTL-based caching and inflight request coalescing
• ✅ Parallelized independent API calls via Promise.all
• ✅ Smart cache TTLs:
  • Trending topics: 5 min
  • Community data: 10 min
  • Feed price / global props: 1 min
  • Posts: 30–60 sec
• Expected improvement: SSR latency ~300ms → ~80–120ms

PR #3973 – SSR Cache Fix

• ✅ Fixed specialPosts() to use cached promise (refreshed every 5 min)
• ✅ Eliminated 10+ redundant blockchain API calls per SSR request

---

Current Status

✅ All four fixes have been merged and deployed to production.
✅ System performance is stable through the previously affected time window.
✅ Continued monitoring in place.

---

What This Means for Users

• Faster page loads – Especially for trending, feed, and community pages
• More responsive SSR – Reduced server-side rendering latency by ~60–70%
• Better resource utilization – Lower CPU usage allows handling traffic peaks more gracefully

---

Thank You

We appreciate the community's patience and feedback during this period. If you experience any further issues, please report them via our standard support channels.