Hi,
We are developing a multithreaded application
which uses the following environment
- Netscape LDAP SDK (version 3.1)
- ACE
- Red Hat Linux 6.1
The threads are daemon threads spawned
at startup and killed when the application terminates.
The number of threads are configurable at start up.
Every thread has its own LDAP*, so we are not
using the thread callback functions provided by the SDK
used in a situation where multiple threads share the same
ldap session .
Every thread initializes its LDAP* using the ldap_init
call. We are using the LDAP_OPT_RECONNECT
to rebind to the server if it goes down.
We are checking for the return value of ldap_search_ext_s().
If the return code is LDAP_SERVER_DOWN or
LDAP_CONNECT_ERROR ,which indicates the server is
down, we rebind to the server using the LDAP* obtained
from the ldap_init call. If the bind succeeds we fire the search again.
In order to simulate a case in which the server goes down, while
performng a search and comes up again, we manually shutdown
the Netscape directory server, while the application fired searches on it.
In some situations we encountered a broken pipe error
(due to SIGPIPE signal) which caused the application to exit.
We observed the same problem even when the application
has just a single thread running.
Has anyone encountered the SIGPIPE signal with the
LDAP_OPT_RECONNECT option?
Should we write a signal handler to catch the signal?
