Management summary · Q3 2026 KPI

Consolidating 4 App Platform apps onto one shared droplet cuts hosting spend by 58%

All four Next.js applications move from DigitalOcean App Platform to a single 8 vCPU / 16 GiB droplet in the same region (BLR1), fronted by Caddy and Docker Compose. App Platform stays live in parallel until each app is verified — cutover is a DNS change per app, with instant rollback.

Apps hosting today

$345/mo

App Platform, 4 apps + egress IP

Apps hosting after

$115/mo

Droplet $96 + weekly backups $19.20

Monthly saving

−$229.80

−58% on app compute

Annualised saving

~$2,758/yr

Before ops-time considerations

Total monthly infrastructure, by phase (includes unchanged cPanel droplet, $48/mo)

Today$393.00 / mo

Migration overlap (1–2 months, both stacks live)$508.20 / mo

Steady state (after App Platform decommission)$163.20 / mo

App Platform New shared droplet cPanel droplet (unchanged)

01 · Architecture

From four managed apps to one owned stack

Databases and storage are untouched — Supabase, Shopify, Cloudinary and R2 stay exactly where they are. Only the compute layer changes.

Today — App Platform

$345/mo

Target

After — Shared droplet

$115.20/mo

What changes	App Platform (today)	Shared droplet (target)
Monthly app compute	$345 — isolated per app	$115.20 — shared
Compute	26 GiB RAM summed across 4 bills; autotraders alone has 4 dedicated vCPU	8 shared vCPU / 16 GiB pooled across all 4 apps
Deploys, SSL, scaling	Fully managed by DigitalOcean	Self-managed: Docker, Caddy, GitHub Actions
elf dedicated egress IP	+$25/mo add-on	Droplet reserved IP — no extra fee
Rollback	DO deployment history	Image tags + DNS revert (App Platform kept live during soak)
Ops burden	Low	Medium — in-house or contractor

02 · Cost analysis

Where the money goes today, and after

Current App Platform spend, per app

App	Instance	$/mo
autotraders	4 dedicated vCPU / 16 GiB	196.00
elf-bangladesh	2 shared vCPU / 4 GiB	50.00
japanpartsbd	1 dedicated vCPU / 4 GiB	49.00
asianimportsltd	1 shared vCPU / 2 GiB	25.00
elf egress IP	Dedicated IP add-on	25.00
Subtotal		345.00

One app drives 57% of the bill

autotraders' dedicated 4-vCPU instance is the single biggest line item consolidation removes.

Droplet sizing options evaluated (incl. 20% weekly backups)

Option	Spec	$/mo
Budget OOM risk	4 vCPU / 8 GiB	57.60
Recommended ✓	Basic 8 vCPU / 16 GiB · 320 GiB disk · 6 TB transfer	115.20
Dedicated CPU	General Purpose 8 vCPU / 16 GiB	151.20
More RAM	8 vCPU / 32 GiB	230.40

The 8 GiB option saves the most but is not recommended — four Next.js apps are likely to run out of memory under load. Builds run on GitHub Actions (within the free tier), never on the droplet.

Any option beats $345/mo

Even the most generous sizing (32 GiB) is a third cheaper than today's App Platform subtotal.

Total infrastructure by phase

Overlap is temporary insurance: both stacks run in parallel for 1–2 months (≈ $173 extra for a 6-week overlap), then spend settles at $163.20/mo.

All figures USD, DigitalOcean list prices

Payback on the overlap period is under one month

A 6-week parallel run costs roughly $173 extra; steady-state savings of $229.80/mo recover that in the first month after decommission.

−58% app compute

03 · Migration status & cutover plan

Staging is already live — production cutover is a DNS change

Completed

✓Droplet provisioned in BLR1 with firewall, backups, Docker
✓All 4 environment configs exported and corrected (URLs, debug flags)
✓All 4 apps containerised and running behind Caddy
✓Internal smoke tests: all 4 backends returning HTTP 200
✓8 build/runtime issues found and fixed during first deploy

Remaining

→Add staging DNS records in Cloudflare; verify public HTTPS
→3–7 day soak on subdomains, comparing against App Platform
→GitHub Actions CI/CD (build in cloud, droplet only pulls images)
→Phased production DNS cutover, one app at a time
→Decommission App Platform after 7–14 day production soak

Cutover order — lowest risk first

Each step is reversible by reverting a single DNS record. App Platform keeps serving each app until its cutover is verified.

STEP 1

elf

New subdomain added — no existing traffic at risk

STEP 2

asianimports

New subdomain added — no existing traffic at risk

STEP 3

japanparts.com.bd

Production domain + Shopify webhook update

STEP 4

autotraderbd.com

Highest traffic; Cloudflare cache purge + bypass rules for admin/auth paths

Key risks and how they're covered

Risk	Mitigation
Downtime during cutover	Parallel run, low DNS TTL, one app at a time, instant DNS rollback
Out-of-memory during builds	Images built on GitHub Actions, never on the droplet
One app destabilising others	Per-container memory limits in Docker Compose
Stale/cached private pages (prior incident)	Cloudflare cache purge + bypass rules on admin, auth and dashboard paths at cutover
Shopify webhooks breaking	Webhook URLs updated before the japanparts cutover
Secret exposure	Runtime secrets live only on the droplet (chmod 600); exposed token flagged for rotation

Out of scope in all scenarios: the existing cPanel/WHM droplet (Singapore, $48/mo) is untouched, and external services — Supabase, Cloudflare, Cloudinary, R2, Shopify — are unchanged either way.