apache/hadoop · error · HttpGetFailedException

Fetch of url failed with status code connection.getResponseC

Error message

Fetch of url failed with status code connection.getResponseCode()\nResponse message:\nconnection.getResponseMessage()

What it means

HttpGetFailedException thrown by the URLLog fetcher inside EditLogFileInputStream: a remote edit log is opened over HTTP (e.g., the EditLogTailer on a standby/observer NameNode pulling segments from a peer's GetImageServlet), and connection.getResponseCode() != 200. The message carries the exact URL, status code, and the server's response message, which identifies the real cause.

Source

Thrown at hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/EditLogFileInputStream.java:494

      this.url = url;
    }

    @Override
    public InputStream getInputStream() throws IOException {
      return SecurityUtil.doAsCurrentUser(
          new PrivilegedExceptionAction<InputStream>() {
            @Override
            public InputStream run() throws IOException {
              HttpURLConnection connection;
              try {
                connection = (HttpURLConnection)
                    connectionFactory.openConnection(url, isSpnegoEnabled);
              } catch (AuthenticationException e) {
                throw new IOException(e);
              }
              
              if (connection.getResponseCode() != HttpURLConnection.HTTP_OK) {
                throw new HttpGetFailedException(
                    "Fetch of " + url +
                    " failed with status code " + connection.getResponseCode() +
                    "\nResponse message:\n" + connection.getResponseMessage(),
                    connection);
              }
        
              String contentLength = connection.getHeaderField(CONTENT_LENGTH);
              if (contentLength != null) {
                advertisedSize = Long.parseLong(contentLength);
                if (advertisedSize <= 0) {
                  throw new IOException("Invalid " + CONTENT_LENGTH + " header: " +
                      contentLength);
                }
              } else {
                throw new IOException(CONTENT_LENGTH + " header is not provided " +
                                      "by the server when trying to fetch " + url);
              }
        

View on GitHub (pinned to 2add963021)

Solutions

1. Reproduce by hand with the URL from the log: kinit, then curl --negotiate -sv '<getedit URL>' -o /dev/null and read the status/message.
2. 404: confirm the segment exists on the source NN (ls current/ or journal report); if simply not ready yet, the tailer retries on its interval -- check edit-lag metrics.
3. 401/403: fix SPNEGO (keytab present, dfs.namenode.kerberos.principal/SPN HTTP/nn@REALM, krb5.conf, clock skew).
4. 5xx: read the source NN's logs for the servlet failure; fix NN health or version mismatch on GetImageServlet parameters.

Example fix

# before: tailer log shows Fetch of http://nn1:9870/getimage?getedit=1&txid=... failed with status code 404
kinit -kt /etc/security/keytab/nn.keytool nn/nn1@REALM
curl --negotiate -s -o /dev/null -w '%{http_code}\n' 'http://nn1:9870/getimage?getedit=1&txid=1234'
# after: 200 -> segment now available, tailer recovers on next cycle; 403 -> fix SPNEGO

kinit -kt /etc/security/keytab/nn.keytab nn/nn1@REALM
curl --negotiate -s -o /dev/null -w '%{http_code}\n' "<the exact getedit URL from the NN log>"
# expect 200 before the tailer can succeed; 401/403 -> fix Kerberos, 404 -> segment absent

catch (IOException e) {
  if (e instanceof HttpGetFailedException h) {          // or match on message 'Fetch of '
    int code = h.getStatusCode();                        // from the exception's connection
    if (code == 404) { scheduleRetry(backoff); }         // segment not ready / purged
    else if (code == 401 || code == 403) { alert("SPNEGO misconfigured"); }
    else { retryWithBackoff(); }
  } else { throw e; }
}

When it happens

Trigger: Opening EditLogFileInputStream on an http(s) URL to the journal endpoint and getting 404 (segment not yet available on that NN or already purged), 401/403 (SPNEGO/Kerberos failure), or 5xx (servlet error on the source NN). Reachable only from NN-internal tailing or tooling that fetches edits over HTTP.

Common situations: Standby tailer lagging until segments were purged; Kerberos keytab/krb5.conf/SPN (HTTP/host@REALM) misconfigured on the tailing side; dfs.namenode.http-address pointing at the wrong host; a firewall or load balancer in the path returning 502/503.

Related errors

• Image transfer servlet at " + url + " failed with status cod
• "Invalid Content-Length header: " + contentLength
• "Content-Length header is not provided by the server when tr
• Image uploading failed, status: %d, url: %s, message: %s
• Unrecognized value '{}' for dfs.http.policy

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/5e1aa09b86f064fe. Report an issue: GitHub.