Hey there!
I’m trying to deploy my Nuxt frontend on Fly
I keep getting this very same message each time
2025-01-12T22:35:05.470 app[7843ed4a34d028] cdg [info] [ 96.430391] Out of memory: Killed process 648 (node) total-vm:74445100kB, anon-rss:383400kB, file-rss:716kB, shmem-rss:0kB, UID:0 pgtables:1304kB oom_score_adj:0
2025-01-12T22:35:06.350 app[7843ed4a34d028] cdg [info] /bin/bash: line 1: 648 Killed nuxt build
Out of memory
Message says Total VM : 74.4GB (so I’m not sure what is means here, I guess its total VM disk?)
Memory usage : 384MB (above what I see on my own computer, yet not so far, but still far below the 2048MB limit, I might be misunderstanding something, looks to be the case)
I tracked my usage on my Mac Studio, in Docker, with
docker stats frontend --format “table {{.Name}}\t{{.CPUPerc}}\t{{.MemUsage}}\t{{.MemPerc}}”
here is at peak :
Here is my dockerfile
#============ BASE IMAGE ============
FROM oven/bun:1
#============ WORKING DIRECTORY ============
WORKDIR /frontend
#============ NODE ENV ===========
RUN if [ "$NODE_ENV" = "production" ]; then \
echo "\033[32m╔══════════════════════════════╗\033[0m" && \
echo "\033[32m║ 🚀 PRODUCTION 🚀 ║\033[0m" && \
echo "\033[32m╚══════════════════════════════╝\033[0m"; \
else \
echo "\033[35m╔══════════════════════════════╗\033[0m" && \
echo "\033[35m║ 🎮 PLAYGROUND 🎮 ║\033[0m" && \
echo "\033[35m╚══════════════════════════════╝\033[0m"; \
fi
#============ SYSTEM PACKAGES ============
RUN apt-get update && apt-get install -y \
python3 \
make \
g++ \
curl \
procps \
sysstat \
bc \
&& rm -rf /var/lib/apt/lists/* \
&& apt-get clean \
&& echo 'while true; do mem=$(free -m | awk "/Mem:/ {printf \"%.1f%%\", \$3/\$2*100}"); cpu=$(top -bn1 | grep "Cpu(s)" | awk "{printf \"%.1f%%\", \$2}"); tput cup $(tput lines) $(tput cols); printf "\033[31m\033[2D MEM:$mem CPU:$cpu\033[0m"; sleep 1; done' > /usr/local/bin/monitor \
&& chmod +x /usr/local/bin/monitor
#============ APPLICATION SETUP ============
COPY package*.json bun.lockb ./
RUN bun install --frozen-lockfile
COPY . .
#============ ENVIRONMENT SETUP ============
ENV PYTHONUNBUFFERED=1
ENV HOST=${HOST_FRONTEND}
ARG NODE_ENV
ENV NODE_ENV=${NODE_ENV}
#============ NETWORK ============
EXPOSE ${PORT_FRONTEND}
#============ BUILD ============
RUN monitor & \
if [ "$NODE_ENV" = "production" ]; then \
echo "Memory usage before build:" && ps -o pid,rss,command ax | grep bun && \
NODE_ENV=production bun run build && \
echo "Memory usage after build:" && ps -o pid,rss,command ax | grep bun; \
else \
echo "Memory usage before build:" && ps -o pid,rss,command ax | grep bun && \
bun run build && \
echo "Memory usage after build:" && ps -o pid,rss,command ax | grep bun; \
fi
#============ RUNTIME ============
CMD if [ "$NODE_ENV" = "production" ]; then \
echo "Memory usage before start:" && ps -o pid,rss,command ax | grep bun && \
echo "Starting Nuxt production server..." && \
ls -la && \
echo "Content of .output directory:" && \
ls -la .output/ || echo ".output directory not found" && \
echo "Content of .output/server directory:" && \
ls -la .output/server/ || echo ".output/server directory not found" && \
bun run start; \
else \
echo "Starting Nuxt development server..." && \
NITRO_HOST=${NITRO_HOST} NITRO_PORT=${NITRO_PORT} NODE_ENV=${NODE_ENV} HOST=${HOST} PORT=${PORT_FRONTEND} bun run dev; \
fi
here is my fly.toml
docker stats synaptic-frontend --format "table {{.Name}}\t{{.CPUPerc}}\t{{.MemUsage}}\t{{.MemPerc}}"
Based on Docker stats, my frontends hits max 90 MB at build, yet my 2048MB memory limit machine crashes with it..
app = "***************"
name = "***************"
primary_region = "cdg"
[build]
dockerfile = "Dockerfile"
[env]
HOST = "0.0.0.0"
HOST_FRONTEND = "0.0.0.0"
NODE_ENV = "production"
PORT_FRONTEND = "8000"
NITRO_HOST = "0.0.0.0"
NITRO_PORT = "8000"
CORS_ORIGINS = "*"
NUXT_PUBLIC_API_BASE = "***************"
PRIVATE_API_URL = "***************"
PUBLIC_API_URL = "***************"
PYTHONUNBUFFERED = "1"
[http_service] # 🔺 Configuration for the HTTP service
internal_port = 8000 # 🔺 Port the application listens on internally
force_https = true # 🔺 Forces all traffic to use HTTPS
auto_stop_machines = true # 🔺 Automatically stops idle machines to save resources
auto_start_machines = true # 🔺 Automatically starts machines when traffic arrives
min_machines_running = 0 # 🔺 Ensures zero downtime during deployments by keeping at least one instance running
max_unavailable = 0 # 🔺 Maximum number of instances that can be unavailable during updates
processes = ["app"] # 🔺 List of processes to run
[[http_service.checks]] # 🔺 Health check configuration
interval = "60s" # 🔺 How often to run health checks
timeout = "20s" # 🔺 Maximum time to wait for health check response
grace_period = "300s" # 🔺 5min for build - Initial delay before starting health checks
method = "GET" # 🔺 HTTP method to use for health checks
path = "/api/health" # 🔺 Endpoint to check for health status
protocol = "http" # 🔺 Protocol to use for health checks
[http_service.concurrency] # 🔺 Concurrency settings
type = "connections" # 🔺 Type of concurrency limit
hard_limit = 500 # 🔺 Maximum number of concurrent connections
soft_limit = 400 # 🔺 Soft limit for concurrent connections
[[vm]] # 🔺 Virtual machine configuration
cpu_kind = "shared" # 🔺 Type of CPU allocation
cpus = 2 # 🔺 Number of CPU cores
memory_mb = 2048 # 🔺 Amount of memory in MB
min_machines_running = 1 # 🔺 Minimum number of VMs to keep running
[deploy] # 🔺 Deployment configuration
strategy = "rolling" # 🔺 Rolling deployment strategy for zero downtime
release_command = "bun run build" # 🔺 Command to run during deployment
[mounts] # 🔺 Volume mount configuration
source = "nuxt_build" # 🔺 Name of the volume to mount
destination = "/frontend/.nuxt" # 🔺 Where to mount the volume in the container
[[statics]] # 🔺 Static file serving configuration
guest_path = "/frontend/public" # 🔺 Path to static files in container
url_prefix = "/public" # 🔺 URL prefix for serving static files
This is my very first time trying to deploy an app , I’ve been asking myself a lot if I should not just use a direct VPS, but fly seems really nice, and pretty good for what I’m doing, hence I wanna fix it instead.
Thanks a lot for any person taking the time to read me.
So, why would I hit the memory limit, if all indicates that I am faaar below it?
I have tried multiple variants of my fly.toml configuration at this point, did multiple deploys , nothing seems to fix it, coming here to seek some help
