What causes "getClassPath_helper: failed strdup: %s " in apache/hadoop?

Raise memory limits or free memory before the first libhdfs call (JVM init is lazy)

getClassPath_helper: failed strdup: %s

Error message

getClassPath_helper: failed strdup: %s

What it means

getClassPath_helper begins by duplicating the CLASSPATH value with strdup so it can tokenize it with strtok. If the duplication fails, 'failed strdup: ...' is printed with strerror(errno) and -1 is returned, aborting classpath preparation and therefore JVM creation on the first libhdfs call.

Source

Thrown at hadoop-hdfs-project/hadoop-hdfs-native-client/src/main/native/libhdfs/jni_helper.c:501

/**
 * Helper to expand classpaths. Returns the total length of the expanded
 * classpath. If expandedClasspath is not NULL, then fills that with the
 * expanded classpath. It assumes that expandedClasspath is of correct size, eg
 * allocated after using this function with expandedClasspath=NULL to get the
 * right size.
 */
static ssize_t getClassPath_helper(const char *classpath, char* expandedClasspath)
{
    ssize_t length;
    ssize_t retval;
    char* expandedCP_curr;
    char* cp_token;
    char* classpath_dup;

    classpath_dup = strdup(classpath);
    if (classpath_dup == NULL) {
        fprintf(stderr, "getClassPath_helper: failed strdup: %s\n",
          strerror(errno));
        return -1;
    }

    length = 0;

    // expandedCP_curr is the current pointer
    expandedCP_curr = expandedClasspath;

    cp_token = strtok(classpath_dup, PATH_SEPARATOR_STR);
    while (cp_token != NULL) {
        size_t tokenlen;

#ifdef _LIBHDFS_JNI_HELPER_DEBUGGING_ON_
        printf("%s\n", cp_token);
#endif

        tokenlen = strlen(cp_token);

View on GitHub (pinned to 2add963021)

Solutions

Raise memory limits or free memory before the first libhdfs call (JVM init is lazy)
Trim the CLASSPATH to only the required jars

Defensive patterns

Strategy: fallback

Validate before calling

/* fail fast when the environment is oversized or memory is scarce */
const char *cp = getenv("CLASSPATH");
if (cp && strlen(cp) > 1 * 1024 * 1024) { /* trim before JVM init */ }

Try / catch

if (hdfsConnect(uri, user) == NULL && errno == EINTERNAL) {
    /* check stderr for 'failed strdup': free memory or shorten CLASSPATH, then retry once */
}

Prevention

Keep CLASSPATH minimal and versioned
Provision enough memory before JVM bootstrap, which happens lazily on first libhdfs call

When it happens

Trigger: Process heap exhausted at JVM-initialization time; an extremely large CLASSPATH string combined with tight memory limits.

Common situations: Machine-generated multi-megabyte CLASSPATH variables; memory-capped containers starting processes that use libhdfs.

Related errors

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/16906b22230f6201. Report an issue: GitHub.