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

Re: MirrorMode Replication Problem

On Tue, Nov 27, 2012 at 07:09:35PM +0900, rocketdive5@yahoo.co.jp wrote:

> I would like to know whether there is the API or plugin which can detect what was replicated.

Not directly, though you could turn on 'stats' and 'stats2' logging
which would show you the entries supplied by one server to another. If
you have problems to diagnose then you will want at least that level of
logging anyway.

It may be easier to find out what *was not* replicated. This
script will show you the current state of each server:

# check-replication

export PATH


for host in host1.example.com host2.example.com host3.example.com ...
        echo $host
        ldapsearch -LLL -x -H ldap://$host/ -b ${SUFFIX} -s base contextCSN

If everything is working then the set of contextCSN values you
see should be the same on each server. The values look like this:

contextCSN: 20120127155755.195827Z#000000#001#000000
contextCSN: 20121129113523.045520Z#000000#003#000000

You can see an obvious date field at the start. The small
integer near the end is the server ID of the server that
initially received the update.

Given a contextCSN value you can find the entry that it relates
to like this:

ldapsearch -x -H ldap://host1.example.com/ "(entryCSN=20121127112456.206198Z#000000#003#000000)"

You might want to try that against each server in the ring just
to check that they all show you the latest entry.

If one server is lagging behind you will see an older
datestamp in the contextCSN. You can then find the entries that
it has not updated by querying the source server, something
like this:

ldapsearch -x -H ldap://host2.example.com/ "(entryCSN>=20121127112456.206198Z#000000#003#000000)"

Note that I am making an assumption here that replication
worked up to a given time and then stopped. This process will
not tell you much about changes that got skipped due to bugs.

Before getting into this sort of detail I suggest you check
that each server is actually making successful connections to
those one each side when it comes up. Also, as Quanah says,
make sure you are running a release that has the fewest
possible number of known replication bugs.

|                 From Andrew Findlay, Skills 1st Ltd                 |
| Consultant in large-scale systems, networks, and directory services |
|     http://www.skills-1st.co.uk/                +44 1628 782565     |