Streaming and chunked transfer in Nginx - Step-by-Step Execution

Process Flow - Streaming and chunked transfer

Client sends HTTP request

↓

Nginx receives request

↓

Nginx starts processing response

↓

Response data generated in chunks

↓

Nginx sends each chunk immediately

↓

Client receives chunks progressively

↓

Last chunk sent with zero size

↓

Client knows response is complete

Nginx receives a request and sends the response in small pieces called chunks immediately as they are ready, letting the client start using data before the full response is done.

Execution Sample

Nginx

location /stream {
  proxy_pass http://backend;
  proxy_buffering off;
  chunked_transfer_encoding on;
}

Nginx config to stream response from backend to client using chunked transfer encoding without buffering.

Process Table

Step	Action	Nginx State	Data Sent to Client	Notes
1	Client sends request to /stream	Waiting for backend response	None	Request received, no data sent yet
2	Nginx forwards request to backend	Waiting for backend response	None	Proxying request
3	Backend starts sending first chunk	Received first chunk	Chunk 1 data	Nginx sends chunk immediately
4	Backend sends second chunk	Received second chunk	Chunk 1 data + Chunk 2 data	Streaming continues
5	Backend sends third chunk	Received third chunk	Chunks 1-3 data	Data sent progressively
6	Backend sends last zero-length chunk	Received last chunk	All chunks + zero-length chunk	Indicates end of response
7	Nginx closes connection	Response complete	Full response received by client	Streaming finished

💡 Last zero-length chunk signals end of streaming, so nginx stops sending data.

Status Tracker

Variable	Start	After Step 3	After Step 4	After Step 5	After Step 6	Final
Nginx buffer	empty	Chunk 1 data	Chunk 1+2 data	Chunk 1+2+3 data	Full data + zero-length chunk	empty after send
Client received data	none	Chunk 1 data	Chunk 1+2 data	Chunk 1+2+3 data	Full response	Full response

Key Moments - 3 Insights

Why does nginx send data in chunks instead of waiting for the full response?

Because proxy_buffering is off and chunked_transfer_encoding is on, nginx streams data immediately to reduce delay, as shown in steps 3-5 where chunks are sent progressively.

How does the client know when the response is complete?

The client detects the last zero-length chunk sent in step 6, which signals the end of the streamed response.

What happens if proxy_buffering is on instead?

Nginx would wait to collect more data before sending, so chunks would not be sent immediately, delaying client reception (not shown in this trace).

Visual Quiz - 3 Questions

Look at the execution table, what data has the client received after step 4?

A Chunk 1 data only

B Chunk 1 and Chunk 2 data

C All chunks including last zero-length chunk

D No data yet

💡 Hint

Check the 'Data Sent to Client' column at step 4 in the execution table.

At which step does nginx send the last zero-length chunk to signal the end of the response?

A Step 6

B Step 5

C Step 3

D Step 7

💡 Hint

Look for the step mentioning 'last zero-length chunk' in the 'Action' column.

If proxy_buffering was enabled, how would the 'Data Sent to Client' column change at step 3?

A Chunk 1 data would be sent immediately

B All chunks would be sent at once

C No data would be sent yet

D Zero-length chunk sent first

💡 Hint

Recall that proxy_buffering off allows immediate sending; enabling it delays sending.

Concept Snapshot

Streaming and chunked transfer in nginx:
- proxy_buffering off disables buffering
- chunked_transfer_encoding on sends data in chunks
- Nginx forwards chunks immediately as backend sends
- Client receives data progressively
- Last zero-length chunk signals end of response

Full Transcript

When a client requests data from nginx configured for streaming and chunked transfer, nginx forwards the request to the backend server. As the backend generates data, it sends it in small pieces called chunks. Nginx immediately forwards each chunk to the client without waiting for the full response, because proxy_buffering is turned off and chunked_transfer_encoding is enabled. The client receives these chunks progressively, allowing it to start processing data sooner. When the backend finishes, it sends a last zero-length chunk, which nginx forwards to signal the end of the response. This method reduces latency and improves user experience for large or slow responses.