Hello people,
My scenario is:
1) I have two LDAP servers, primary and slave.
1) LDAP servers are told to close idle connections within the
given timeout.
2) Exim uses those two servers in a 4.12 "chained" way,
"this-fail-ask-next".
I observe periodical mail recipient rejection, especially on large
envelopes with many recipients. As I can see from the code, Exim tries
to use cached LDAP connections, but if one is closed from the
server-side, ldap_errno() is obviously LDAP_SERVER_DOWN. Then Exim
tries to switch to the next available server. If Exim gets stucked
again with DNS lookups/etc, my slave server also closes idle
connection. Again, Exim gets LDAP_SERVER_DOWN on a last available
server, and gives up with the rest of recipients.
Opposing to MySQL, LDAP client API does not have any kind of automatic
reconnection (rebinding) logic triggered by flags/etc. That's why this
lookup issue seems to be LDAP-specific.
I think that Exim LDAP connection caching/binding logic should be
reviewed in such a way that Exim should try to unbind/connect/bind
cached connection once before giving up with the server. This gives
some overhead when server is just gone.
What do You think?
2 Philip: no problem, I can handle it. Just asking before coding ;-)
--
Best regards,
Peter mailto:spam4octan@highway.ru
