What memcache version do you use?
Please provide GDB backtrace also.

pecl-memcache-2.0.4
memcached-1.2.0

Backtrace was generated by pecl-xdebug(I think) in another script that uses pconnect.

*** glibc detected *** /usr/sbin/apache2: munmap_chunk(): invalid pointer: 0x000000000086b3a8 ***
======= Backtrace: =========
/lib/libc.so.6(__libc_free+0x174)[0x2ba5b2698474]
/usr/lib64/php5/lib/php/extensions/no-debug-non-zts-20050922/memcache.so[0x2ba5b7334d0d]
/usr/lib64/php5/lib/php/extensions/no-debug-non-zts-20050922/memcache.so[0x2ba5b733648a]
/usr/lib64/php5/lib/php/extensions/no-debug-non-zts-20050922/xdebug.so(xdebug_execute_internal+0x154)[0x2ba5b7208a0b]
/usr/lib64/apache2/modules/libphp5.so[0x2ba5b45b1498]
/usr/lib64/apache2/modules/libphp5.so(execute+0x173)[0x2ba5b45a1673]
/usr/lib64/php5/lib/php/extensions/no-debug-non-zts-20050922/xdebug.so(xdebug_execute+0xac7)[0x2ba5b72086cd]
/usr/lib64/apache2/modules/libphp5.so[0x2ba5b45a7216]
/usr/lib64/apache2/modules/libphp5.so(execute+0x173)[0x2ba5b45a1673]
/usr/lib64/php5/lib/php/extensions/no-debug-non-zts-20050922/xdebug.so(xdebug_execute+0xac7)[0x2ba5b72086cd]
/usr/lib64/apache2/modules/libphp5.so(zend_execute_scripts+0x16f)[0x2ba5b4584c3f]
/usr/lib64/apache2/modules/libphp5.so(php_execute_script+0x229)[0x2ba5b45481f9]
/usr/lib64/apache2/modules/libphp5.so[0x2ba5b460117d]
/usr/sbin/apache2(ap_run_handler+0x79)[0x425c89]
/usr/sbin/apache2(ap_invoke_handler+0x72)[0x428b52]
/usr/sbin/apache2(ap_process_request+0x158)[0x423c58]
/usr/sbin/apache2(apr_vformatter+0x9c0)[0x41ec60]
/usr/sbin/apache2(ap_run_process_connection+0x71)[0x42fb91]
/usr/sbin/apache2[0x424973]
/usr/sbin/apache2[0x424b94]
/usr/sbin/apache2(ap_mpm_run+0x647)[0x425297]
/usr/sbin/apache2(main+0x5ed)[0x42b2ed]
/lib/libc.so.6(__libc_start_main+0xf4)[0x2ba5b2647134]
/usr/sbin/apache2(apr_bucket_mmap_make+0x59)[0x41e5b9]

Please update to the latest stable version first (2.1.0) and if you still can replicate it, provide a backtrace as described here: 
http://bugs.php.net/bugs-generating-backtrace.php

>To get a backtrace with correct information you must have PHP configured with --enable-debug

I am sorry, can't do that. I have no time for that. And there is no ebuild in gentoo for 2.1.0 :(

I can't reproduce it too, so I guess it's fixed.
Feel free to reopen the report when/if you have more information to tell.

I do believe I experience this or an similar bug with 3.0.2 and memcached 1.2.6.

I have an job that runs as an CLI-job. This job multi-gets about 45.000-50.000 keys and loops through all of them and does some delete and some sets depending on the information but mostly it does nothing. This multi-get normally takes nothing more then a few tenths of a second.

The problem occurs sometimes when the memcached-cluster is under pretty high load(~10000 request per second per server(14 servers)). Now the multi-get suddenly take 4-8 minutes to complete and all the following deletes and sets Memcached is supposed to have done isn't successfully.

The job finishes(I have an echo on the last line) and after that I get this:

*** glibc detected *** corrupted double-linked list: 0x0000000002a09260 ***
Aborted

All sets and deletes returns FALSE. The servers are added to the connection pool using addServer().