How to read a file stream in Node.js

Large text files are safer to process incrementally when a Node.js job needs only a bounded amount of data in memory at once. A readable file stream supplies chunks as the file system makes them available, which suits log scans and import jobs.

The fs.createReadStream() method returns a readable stream, and async iteration consumes it to end-of-file while propagating a read failure to the surrounding try / catch block. Setting encoding to utf8 makes each chunk a string instead of a Buffer.

A chunk can end halfway through a line, so the parser must carry the unfinished suffix into the next chunk. The sample lowers highWaterMark to 32 bytes only to expose that boundary; application code should keep the default unless measurements justify tuning it.

1. Create the sample log file as events.log.

   events.log

   2026-06-28T09:00:00Z INFO stream opened
   2026-06-28T09:00:01Z INFO chunk accepted
   2026-06-28T09:00:02Z ERROR payment retry queued
   2026-06-28T09:00:03Z INFO stream closed

2. Create read-stream.mjs with the stream configuration and counters.

   read-stream.mjs

   import { createReadStream } from "node:fs";
    
   const filePath = process.argv[2] ?? "events.log";
   const stream = createReadStream(filePath, {
     encoding: "utf8",
     highWaterMark: 32,
   });
    
   let chunks = 0;
   let bytes = 0;
   let lines = 0;
   let errorLines = 0;
   let carry = "";

3. Append the record-counting helper to read-stream.mjs.

   read-stream.mjs

   function countRecord(record) {
     if (record === "") return;
    
     lines += 1;
     if (record.includes("ERROR")) {
       errorLines += 1;
     }
   }

4. Append the chunk loop after countRecord().

   read-stream.mjs

   try {
     for await (const chunk of stream) {
       chunks += 1;
       bytes += Buffer.byteLength(chunk, "utf8");
    
       const records = (carry + chunk).split("\n");
       carry = records.pop() ?? "";
    
       for (const record of records) {
         countRecord(record);
       }
     }

5. Append the completion and error-handling section below the chunk loop.

   read-stream.mjs

     countRecord(carry);
    
     console.log(`file=${filePath}`);
     console.log(`chunks=${chunks}`);
     console.log(`bytes=${bytes}`);
     console.log(`lines=${lines}`);
     console.log(`errorLines=${errorLines}`);
   } catch (error) {
     console.error(`Failed to read ${filePath}: ${error.code ?? error.message}`);
     process.exitCode = 1;
   }

   The try / catch block receives stream errors from async iteration and leaves the process with exit status 1 after a failed read.

6. Compare the assembled read-stream.mjs with the complete source.

   read-stream.mjs

   import { createReadStream } from "node:fs";
    
   const filePath = process.argv[2] ?? "events.log";
   const stream = createReadStream(filePath, {
     encoding: "utf8",
     highWaterMark: 32,
   });
    
   let chunks = 0;
   let bytes = 0;
   let lines = 0;
   let errorLines = 0;
   let carry = "";
    
   function countRecord(record) {
     if (record === "") return;
    
     lines += 1;
     if (record.includes("ERROR")) {
       errorLines += 1;
     }
   }
    
   try {
     for await (const chunk of stream) {
       chunks += 1;
       bytes += Buffer.byteLength(chunk, "utf8");
    
       const records = (carry + chunk).split("\n");
       carry = records.pop() ?? "";
    
       for (const record of records) {
         countRecord(record);
       }
     }
    
     countRecord(carry);
    
     console.log(`file=${filePath}`);
     console.log(`chunks=${chunks}`);
     console.log(`bytes=${bytes}`);
     console.log(`lines=${lines}`);
     console.log(`errorLines=${errorLines}`);
   } catch (error) {
     console.error(`Failed to read ${filePath}: ${error.code ?? error.message}`);
     process.exitCode = 1;
   }

7. Check the missing-file error path.

   $ node read-stream.mjs missing.log
   Failed to read missing.log: ENOENT

8. Run the stream reader against events.log.

   $ node read-stream.mjs events.log
   file=events.log
   chunks=6
   bytes=169
   lines=4
   errorLines=1