(notice) child pid XXXX exit signal Segmentation fault (11), possible coredump in /etc/apache2

(notice) child pid XXXX exit signal Segmentation fault (11), possible coredump in /etc/apache2

  1. Blog
  2. 15 November 2026
  3. 4 min read

Stefan Bogdanescu

Founder & Senior Architect

Decoding Apache Segmentation Faults in a Shared Environment

Seeing [notice] child pid XXXX exit signal Segmentation fault (11) in your Apache logs is immediately alarming, but the log entry itself often provides more confusion than clarity. The message suggests a core dump might exist in /etc/apache2, yet finding and analyzing that file is rarely productive for pinpointing the exact line of code causing the crash. When you are dealing with a production server hosting WordPress on Ubuntu, your primary goal should be isolating the trigger without diving into low-level core dumps unless absolutely necessary.

The traceback provided in typical debugging attempts often leads to waitpid() and generic internal functions like reader_loop(). This happens because the master Apache process waits for its child workers to finish. When a child dies via SIGSEGV, the parent catches the signal while waiting, resulting in a stack frame that looks like it happened inside the parent's wait logic rather than the actual crash site. Attaching GDB to the main httpd process usually only shows you the state of the master thread, which is why you might see an exit code of 01 (SIGABRT) or just generic function calls without your application symbols.

To actually solve this, you must shift focus from the master process to the specific child process that died. The critical step is retrieving the core dump generated by the crashing child and analyzing it in isolation. Ensure ulimit -c is set to unlimited, as shown in your environment, so that the system captures the memory snapshot. Once the core file is present (often named core.<pid> or simply core), launch GDB specifically against the Apache binary with that core file:

sudo gdb /usr/sbin/apache2 /var/crash/core.7505
(gdb) bt full

The bt full command is essential here because it attempts to load symbols from the shared libraries linked at the time of the crash, not just the current system state. If your GDB session fails to find symbols or exits immediately with "bad user name," it indicates that GDB is trying to access resources restricted by the APACHE_RUN_USER directive. You must launch GDB as root (using sudo) to bypass these permission checks when attaching to a crashed process owned by the Apache user.

A common, often overlooked cause for segmentation faults in this specific configuration involves the interaction between modules and PHP's memory management. Your environment includes Suhosin-Patch, APC, mod_remoteip, and Google PageSpeed Service. Each of these introduces distinct risk vectors:

  1. mod_remoteip: This module rewrites the X-Forwarded-For header. If the remote IP address provided by the load balancer or proxy is malformed, excessively long, or contains control characters, PHP's underlying C extensions might crash when parsing the header. Since this happens before APC caching takes over, it suggests the issue occurs during the initial request handling.
  2. APC: While you mentioned installing APC later, if you are rolling back to that state, ensure apc.shm_segments is configured correctly. Improper memory segmentation can lead to crashes when multiple processes attempt to access the same shared segment concurrently.
  3. Google PageSpeed Service: Integration with external services often requires specific headers or timeouts. If the service sends a request that exceeds PHP's execution time limits (max_execution_time) or consumes excessive memory before hitting a timeout, it can trigger a segmentation fault if the cleanup logic isn't robust enough to handle the abrupt termination gracefully.

Another frequent culprit is the version of PHP and its compiled extensions. You are running PHP 5.3.10 on Ubuntu. This version is quite old and may lack fixes for specific memory corruption bugs present in newer releases. If your host environment has been updated to a newer kernel or Glibc while Apache was built against older libraries, incompatibilities can arise that manifest as random segmentation faults.

Finally, consider the Landscape monitoring system errors appearing after installing GDB. The KeyError: 't (tracing stop)' in /var/log/landscape/sysinfo.log indicates that the monitoring agent is reading a process state string that its internal dictionary does not recognize. While this is likely unrelated to the Apache crash, it suggests your server state is being interpreted differently by various tools simultaneously. Ensure that GDB and similar debuggers are not holding file handles or locks that interfere with other system daemons in unexpected ways, although this is rare on modern kernels.

The most effective path forward is to reproduce the crash with a minimal test script rather than relying on the live WordPress site. Strip down the environment: disable mod_remoteip, clear APC segments, and comment out PageSpeed modules. If the segmentation fault disappears, incrementally re-enable components until the specific module or configuration directive causing the instability is identified. Once isolated, upgrading PHP to a supported Long-Term Support (LTS) version or adjusting memory limits (memory_limit in php.ini) often resolves these legacy issues without needing to modify complex core logic.

Our network