Forwarded, as Kiran seems not to be able to send this mail to the ML
---------- Forwarded message ---------- From: Kiran Ayyagari <[email protected]> Date: Mon, Jan 14, 2013 at 9:58 AM Subject: Re: Replica duplicates detection... To: Apache Directory Developers List <[email protected]> On Sun, Jan 13, 2013 at 7:33 PM, Emmanuel Lécharny <[email protected]>wrote: > Hi ! > > what should we do if we get a replication request from a host we already > know, with the same search filter, but an empty cookie ? > > We have two cases here : > > 1) We already have a connected consumer from this host with the given > search filter > It's very likely the consumer has been disconnected, but we haven't yet > noticed the disconnection. However, it may also be another consumer on > the same host... > > Should we consider it's a new consumer ? I bet so... yes > 2) We don't have a connected consumer from this host > Most certainly, the consumer is trying to reconnect. ATM, we create a > new ReplicaEventLog, and we consider that's a new consumer. We keep the > old journal and create a new journal. > > Won't it be better to consider that it's the same consumer ? Should'nt > we reset the old journal? there is no way to detect if it is the same consumer or not, so we must create a new journal/event log > OTOH we *may* have ore than once consumer on the same host, so e may > want to have more than one journal... In any case, we will eventually > detect that the disconnected consumer is dead, and we will remove the > old journal. yup > I do think that in this case, keeping the journal is probably necessary, > assuming we have a mechanism that get rd of it after a period of > inactivity, and without a reconnection from the consumer (ie, a > connection with a valid cookie). yes -- Kiran Ayyagari http://keydap.com
