Implementation:Apache Spark AmIpFilter
| Knowledge Sources | |
|---|---|
| Domains | Security, YARN, Web_UI |
| Last Updated | 2026-02-08 22:00 GMT |
Overview
Servlet filter that validates incoming web requests to the Spark UI originate from authorized YARN proxy hosts and handles redirection.
Description
AmIpFilter is a Jakarta Servlet `Filter` implementation annotated with Hadoop's `@Public` classification. It is copied from Hadoop 3.4.0's `org.apache.hadoop.yarn.server.webproxy.amfilter.AmIpFilter` and modified to use Jakarta Servlet APIs (migrated from javax.servlet). The filter maintains a cached set of proxy IP addresses (refreshed every 5 minutes) and checks each incoming request's remote address against this set. Requests from authorized proxy hosts pass through with the proxy user identity extracted from cookies. Requests from unauthorized sources are redirected to the active YARN ResourceManager proxy URL. It supports RM HA (High Availability) configurations by probing multiple RM URLs.
Usage
This filter is automatically configured in the Spark Web UI servlet container when running on YARN. It ensures that the Application Master's web UI is only accessible through the YARN web proxy, providing security isolation for Spark applications in multi-tenant YARN clusters.
Code Reference
Source Location
- Repository: Apache_Spark
- File: resource-managers/yarn/.../AmIpFilter.java
- Lines: 1-239
Signature
@Public
public class AmIpFilter implements Filter {
public static final String PROXY_HOSTS = "PROXY_HOSTS";
public static final String PROXY_URI_BASES = "PROXY_URI_BASES";
public static final String PROXY_USER_COOKIE_NAME = "proxy-user";
@Override
public void init(FilterConfig conf) throws ServletException { ... }
@Override
public void doFilter(ServletRequest req, ServletResponse resp,
FilterChain chain) throws IOException, ServletException { ... }
protected Set<String> getProxyAddresses() throws ServletException { ... }
public String findRedirectUrl() throws ServletException { ... }
public boolean isValidUrl(String url) { ... }
}
Import
import org.apache.spark.deploy.yarn.AmIpFilter;
I/O Contract
Inputs
| Name | Type | Required | Description |
|---|---|---|---|
| PROXY_HOSTS | FilterConfig param | Yes | Comma-separated list of proxy hostnames |
| PROXY_URI_BASES | FilterConfig param | Yes | Comma-separated list of proxy base URIs |
| RM_HA_URLS | FilterConfig param | No | Comma-separated RM HA URLs for failover |
| HttpServletRequest | Jakarta Servlet | Yes | Incoming HTTP request to filter |
Outputs
| Name | Type | Description |
|---|---|---|
| Pass-through | FilterChain | Authorized requests continue with proxy user principal set |
| Redirect | HTTP 302 | Unauthorized requests redirect to YARN proxy URL |
Usage Examples
Automatic YARN Integration
// AmIpFilter is typically configured automatically by Spark's YARN integration.
// Manual configuration in web.xml (for reference):
//
// <filter>
// <filter-name>AmIpFilter</filter-name>
// <filter-class>org.apache.spark.deploy.yarn.AmIpFilter</filter-class>
// <init-param>
// <param-name>PROXY_HOSTS</param-name>
// <param-value>proxy1.example.com,proxy2.example.com</param-value>
// </init-param>
// <init-param>
// <param-name>PROXY_URI_BASES</param-name>
// <param-value>http://proxy1.example.com:8088/proxy/app_123,
// http://proxy2.example.com:8088/proxy/app_123</param-value>
// </init-param>
// </filter>