Slide 1 — Claude Writes a Lot

Docker images, log files, node_modules copies, temporary build artifacts. Claude generates code — but cleaning up is on you. These commands show you immediately what is eating your server:

# How full is the disk?
df -h

# Which directory is consuming the most?
du -sh /* 2>/dev/null | sort -rh | head -10

# Kill Docker leftovers (images, containers, volumes, cache)
docker system prune -af --volumes

# How large is a single directory?
du -sh /opt/techlogia/

Rule of thumb: When df -h shows over 80%, things will break soon.

Slide 2 — Server Full

The most common reason deploys silently die: no disk space left. Logs grow to gigabyte size without anyone noticing.

# Find largest files on the system
find / -type f -size +100M -printf '%s\t%p\n' 2>/dev/null | sort -rn | head -10

# Empty a log file without deleting it
truncate -s 0 /var/log/nginx/access.log

# Limit systemd logs
journalctl --vacuum-size=200M

Slide 3 — Port Already in Use

Claude starts the dev server — but port 3000 is still running from last time. Instead of guessing:

# Who is listening on port 3000?
lsof -i :3000

# Show all occupied ports
ss -tlnp

# Kill the process on port 3000
kill -9 $(lsof -t -i :3000)

# Or more directly:
fuser -k 3000/tcp

Slide 4 — Keep Processes Under Control

After 10 Claude sessions, sometimes 5 instances of the same service are running. Here is how to keep track:

# All running processes sorted by memory usage
ps aux --sort=-%mem | head -15

# Interactive process monitor
htop

# Kill process by name
pkill -f "node server.js"

# Kill all processes of a user
pkill -u www-data

Slide 5 — Read Logs Without Drowning

Claude outputs an error — but which log file? And what happened just before?

# Last 50 lines of a log
tail -n 50 /var/log/nginx/error.log

# Watch live
tail -f /var/log/syslog

# Search for errors in Docker container logs
docker logs techlogia-backend --tail 100 | grep -i error

# Logs since a specific time
journalctl --since "2026-04-23 10:00" --until "2026-04-23 11:00"

Slide 6 — Find the Mess: Claude Created 47 Files

Not sure what Claude created anymore? These commands show you:

# Files sorted by modification date (newest first)
ls -lt /opt/myproject/ | head -20

# All files newer than 1 hour
find . -newer /tmp/timestamp -type f 2>/dev/null

# Set a timestamp (then see all files created after it)
touch /tmp/timestamp

# How many files are in a directory?
find . -type f | wc -l

# Find duplicate files
# apt install fdupes  (if not yet installed)
fdupes -r . 2>/dev/null | head -20

# Alternative without installation:
find . -type f -exec md5sum {} + 2>/dev/null | sort | awk 'NR>1 && prev==$1 {print p; print $2} {prev=$1; p=$2}' | head -20

Slide 7 — Replace Deploys: No Docker, No CI/CD, No Stress

Sometimes you just want to push a file to the server fast. No pipeline, no workflow:

# Copy a single file to server
scp -i ~/.ssh/mykey file.py root@server:/opt/app/

# Sync a whole directory (changes only)
rsync -avz --delete ./dist/ root@server:/opt/app/dist/

# Run command on server without logging in
ssh root@server "cd /opt/app && git pull && docker compose restart backend"

# Emergency rollback
ssh root@server "cd /opt/app && git revert HEAD --no-edit && docker compose up -d"