[Date Prev][Date Next]
Re: OL 2.3.18 syncrepl vs slurpd
--On Monday, January 23, 2006 1:54 PM -0500 Francis Swasey
On 1/23/06 12:59 PM, Quanah Gibson-Mount wrote:
--On Monday, January 23, 2006 9:04 AM -0500 Francis Swasey
Now, I have discovered three things.
1) delta-syncrepl doesn't seem to have any way to filter the amount of
what is sent -- so, it has the same issues that I'm fighting with slurpd
of sending every update to all the replicas and perhaps I do not want
all the updates on all the replicas (this was the reason for me going to
You could, of course, have more than one accesslog database, each with
what you wanted going to the different replicas. Or, alternately, you
should be able to use a filter similar to what you configured for
syncRepl for use with delta-syncrepl on the accesslog DB.
Yes, I suppose I could, but I don't seem to be intelligent enough to
figure out how to filter what gets put into the accesslog based on
whether or not the DN is in the correct branch of the DIT.
Well, for the filter bit, I meant the logfilter in the replica's syncrepl
configuration. Assuming it is a valid one that doesn't cause the master's
slapd to core dump, you should be able to restrict the branch it looks at
(hint: see the slapcat(8C) manpage for the example of excluding a subtree,
you can change that to require a subtree).
I've probably allowed too much in a single database, and now I want to
send all the updates to one class of replicas and everything except the
sendmail access control entries to another class of replicas -- of
course, this is because slurpd is getting behind from time to time --
perhaps delta-syncrepl will be fast enough that this will not be an issue.
I think delta-syncrepl and slurpd are generally the same speed, and both
are much faster than syncrepl.
Principal Software Developer
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html