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

Re: SyncRepl limitations

Quoting Mark Mcdonald <mmcdonald@staff.iinet.net.au>:


I'm currently writing a report to justify to the powers above that we should upgrade OpenLDAP and move to SyncRepl from slurpd.

A couple of questions:

* Using SyncRepl, what are the time limitations as far as slave refreshing goes? i.e. master & slave are in sync at the moment, can I switch off the slave and turn it on a week later and retrieve all adds, updates & deletes that occurred in the week or is there a 24hr limit? If there's a limit, is it configurable?

I'm not aware of any such limits. Don't remember seeing them mentioned anywhere in documentation.

* Again using SyncRepl, is it possible to create a slave from an empty DB? I.e. I create a new slave with an empty DB, can I just start it and let syncrepl send the entire contents of the database over the wire? I understand this is a great way to congest the network in practice, but I'm only interested in theory at the moment.

It works. You configure syncrepl in slave's slapd.conf, fire it up, and it pulls entire database. Basically, that's how I'm creating my slaves (note, I have relatively small number of entries in directory).

I don't tbink it should kill your network either. Unless you create and fire up 50 new slaves simultaniously and your database has milions of entries ;-) There's probably more bandwith efficient way to transfer database to new slave (dump it, gzip it, than trasfer it), however I found that syncrepl does decent job too.

See Ya' later, alligator!