Re: (ITS#4420) Hanging CLOSE_WAIT connections in ldap-backend

[aar@cpan.org]

I'm experiencing the CLOSE_WAIT problem with slapd 2.3.32 on MacOS X
(PPC). Cached connections aren't being leaked but they aren't closed on
idle-timeout and they remain in CLOSE_WAIT state after the remote host
has gone.

Compile options:
./configure --enable-rewrite --enable-ldap --enable-meta
--enable-proxycache --enable-spasswd

slapd.conf:
===========
database ldap
suffix "dc=AddressBooks,dc=example,dc=com"
uri ldap://localhost:8080
idle-timeout 2
idassert-bind bindmethod=none
    binddn="cn=Administrative Identity,dc=sssss"
    credentials="admin_password"
    authzID="dn:cn=Sandbox,dc=local,dc=example,dc=com"
    mode=self
    authz=proxyAuthz

(The problem appears also without idle-timeout.)

Steps to reproduce:
1. start slapd
2. connect to slapd and do some queries
3. disconnect the proxied server (for example by killing it)
4. netstat -a still shows the connection as CLOSE_WAIT

Debug:
when running slapd with "-d 5", nothing is printed when the connection
is closed by the remote host (i.e. when I kill it), so slapd isn't
probably catching the event at all.

  - alessandro ranellucci.