process still doesn't die + replication stops (ITS#2570)

[Thomas.von_Kamptz@ecb.int]

Full_Name: Thomas-Henning von Kamptz
Version: 2.1.21
OS: SunOS 5.8
URL: ftp://ftp.openldap.org/incoming/
Submission from: (NULL) (195.126.128.2)


We run a setup of secure read only ldap servers beeing fed by one read-write
server.
All systems are now upgraded to 2.1.21 running on SunOS 5.8 with sleepycat db
4.1.25.
* After some time (12-36 hours) one ro-server (the one which is most loaded)
stops picking up changes from the rw-server.
* When this happens this server will not terminate on a simple kill. The picture
is pretty much as already described in #2502.
Kill -9 helps, but the database then keeps some left over locks. 
* After cleanup of those locks, and a restart this server still doesn't pick up
any changes from the master. Here it now looks like slurpd forgot that this link
was configured with ssl:
 conn=81 fd=10 ACCEPT from IP=10.x.y.z:49253 (IP=10.y.z.x:389)
 conn=81 op=0 BIND dn="cn=task-repl,ou=ldapadm,ou=app,dc=ecb,dc=int" method=128
 conn=81 op=0 RESULT tag=97 err=13 text=confidentiality required
 conn=81 op=1 UNBIND
 conn=81 fd=10 closed
* a restart of slurpd fixes that behavior