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

Re: Replication problem with slurpd

Raphaël Ouazana-Sustowski wrote:

I have a problem with slurpd. I have one master which replicates on 2
slaves which one replicates on one other slave.
M -> S1
 -> S/M -> S2
They are all OpenLDAP 2.2.11 (I tested once with 2.2.16 and had the same
problem).

Each night I delete all the entries of the 4 directories and re-populate
the master (M) with ldapadd. So slurpd of the Master populates S1 and
the Slave/Master (S/M) ; and slurpd of S/M populates S2.
The problem is that when I check if all entries are really in the 4
directories, I can see that some are missing (about 0 to 4, rarely
more)! For example I can have :
M : 118821 entries
S1 : 118819
M/S : 118819
S2 : 118818
Of course I have nothing in .rej files, and see nothing in slurpd logs
when I activate them.

Moreother the problem seems to be pretty hard to reproduce. I tried to
reproduce it with test data but don't success. So I'm looking for
methods to analyse the problem. For example how to log properly slurpd ?


See my comment in ITS#3421. As son as you can identify what DNs didn't make it into the slaves, you should first see the reason of the failure at the slapd's level; for this purpose, a loglevel of 256 should suffice: all you need to do is track down what conn/op is related to their operation; they should show up something like

conn=1 op=1 MOD dn="cn=James A Jones 1,ou=Alumni Association,ou=People,o=University of Michigan,c=US"

(ADD/MOD/DEL based on the type of operation they were subjected to at the time of the failure). Similar lines should appear on the master and on the slave; if they don't show up at the slave, this means rplication didn't occur at all; then we need to see if the master says anything at all about that. Later on, you should see lines like

conn=1 op=1 RESULT tag=107 err=0 text=

this can tell you more about te reason of the failre if any. The above, for instance, indicates success (err=0). The "conn=X op=Y" allows to match the operation description with the result. I'm not that familiar with slurpd's log; I see it tends to be very verbose (as slapd's, but I'm more accustomed to selecting the values I need), so I'm afraid you'll have to ask someone else, or experimet yourself. It's my understanding that you can get quite far by looking at the above first, to narrow down the problem.

p.





   SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497