[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Consumer retry with delta-syncrepl

Jan-Piet Mens wrote:
FWIW, if an entry on the provider is modified every two minutes,
the consumer doesn't stop (have been monitoring this for about
seven days). OTOH, an update every five minutes is not sufficient,
and the consumer "falls asleep" and doesn't sync with the provider
any longer.

It sounds to me like one of your intervening routers is aging out and killing the connection without sending a TCP Reset to both sides. Since you're running in refreshAndPersist mode, the consumer normally would keep a connection open to the provider at all times. As such, it won't retry anything unless it actually detects that the connection has closed. You might try letting the 5 minute span pass while running tcpdump to monitor all of the traffic, and see what exactly happens to the syncrepl connection.
On Fri Mar 31 2006 at 08:32:40 CEST, Jan-Piet Mens wrote:

I'm noticing that a slapd consumer (OpenLDAP 2.3.20) located at the
end of an Internet connection with high latency and rather low bandwidth
(somewhere along the lines of 64kbit) isn't retrying connections to
the slapd provider. My configuration is

syncrepl rid=120
        retry="60 10 600 +"

The last 'syncrepl_message_to_op' in slapd'd log is dated 19:00 hours,
(I checked at 07:30 the next morning). Stopping the slapd consumer and
restarting it, causes replication to start up again.

Is there any magic I can apply on the consumer slapd (such as giving it
a signal) to get the replication to kick in again?

Thanks & regards,

 -- Howard Chu
 Chief Architect, Symas Corp.  http://www.symas.com
 Director, Highland Sun        http://highlandsun.com/hyc
 OpenLDAP Core Team            http://www.openldap.org/project/