slapd lockup (ITS#210)

[chris.burrows@bt.com]

Full_Name: Chris Burrows
Version: latest-stable
OS: Solaris 2.6
URL: ftp://ftp.openldap.org/incoming/chris-burrows-990621
Submission from: (NULL) (194.73.73.69)


I have been stress testing the server using the Mindcraft DirectoryMark1.1 tools
and appear to have found a slapd lockup problem where it stops responsing to
network requests.

The test database has approx. 500,000 entries, and the tests follow the
Messaging scenario. This is a large number of requests using a simple indexed
attribute. If I perform the test run 2 or 3 times in sucession, the server
usually locks-up.

Enclosed is a stack back-trace for information.

(gdb) bt
#0  0xef5b62ac in _libc_sigtimedwait ()
#1  0xef5099f0 in _panic ()
#2  0xef5083f0 in _alloc_chunk ()
#3  0xef508010 in _alloc_stack ()
#4  0xef50ebf8 in _alloc_thread ()
#5  0xef50be00 in _thrp_create ()
#6  0xef50fe88 in _pthread_create ()
#7  0x438a0 in ldap_pvt_thread_create (thread=0x6bce870, detach=1,
    start_routine=0x1f088 <connection_operation>, arg=0x6a010b8)
    at thr_posix.c:62
#8  0x1f83c in connection_activity (conn=0xab230) at connection.c:207
#9  0x1eda0 in slapd_daemon (port=0x82000) at daemon.c:362


OS: Solaris 2.6 with recommended patches (from Sunsite 17/6/99)
Compiler: gcc-2.8.1
Berkeley DB: 2.7.5