Re: sync replication questions

[Quanah Gibson-Mount <quanah@zimbra.com>]

--On Wednesday, November 04, 2009 3:03 PM -0500 Edward Capriolo 
<edlinuxguru@gmail.com> wrote:

> All,
> I had an event happen I would like to understand how openldap handled
> this and why.
>
> We have two openldap nodes doing sync replication (configuration is
> below). One node locked up this weekend no response to ping etc. This
> in itself is a problem (since this server is not very high load)
> however I do not have any diagnostics on this so I will move on.
>
> ldap1 crashed. It stayed off for a good part of the weekend. On
> returning to work ldap1 rebooted and openldap was started.
> At this point changes made to ldap2 were propagating to ldap1. However
> changes to ldap1 were not replicating to ldap2.
>
> I restarted ldap1 with only replication debug on. I saw this... (dont
> mind the header upfront we are using daemon tools)
>
> [root@nyldap1 ~]# tail -f /service/openldap/log/main/current
> @400000004af1d1582b3c9c74 bdb_db_open: warning - no DB_CONFIG file found
> in dire ctory /usr/local/openldap/var/openldap-data: (2).
> @400000004af1d1582b3caffc Expect poor performance for suffix "o=ec,c=US".
> @400000004af1d1582b88090c bdb_monitor_db_open: monitoring disabled;
> configure mo nitor database to enable
> @400000004af1d1582b8bfcc4 slapd starting
> @400000004af1d1592bf58c34 slap_client_connect:
> URI=ldap://nyldap1.ops.ec.com DN="cn=root,o=ec,c=us" ldap_sasl_bind_s
> failed (-5)
> @400000004af1d15a00006f54 do_syncrepl: rid=003 rc -5 retrying (4 retries
> left) @400000004af1d15a00022ca4 do_syncrep2: rid=004
> LDAP_RES_INTERMEDIATE - REFRESH_D ELETE
> @400000004af1d15a2aac52f4 TLS: can't accept: (null).
> @400000004af1d15f2ab0f674 TLS: can't accept: (null).
> @400000004af1d15f2ab8f16c do_syncrep2: rid=003 LDAP_RES_INTERMEDIATE -
> REFRESH_D ELETE
> @400000004af1d1642aafae54 TLS: can't accept: (null).
> @400000004af1d1692ab2fa14 TLS: can't accept: (null).
>
> Changes from nyldap1 were still not propagating to nyldap2.
>
> Then I restarted nyldap2. replication was again working in both
> directions.
>
> I based my setup on these notes:
> http://www.linuxtopia.org/online_books//network_administration_guides/lda
> p_administration/replication_LDAP_Sync_Replication.html
>
>
> syncrepl rid=004
>        provider=ldap://nyldap2.ops.ec.com
>        binddn="cn=root,o=ec,c=US"
>        bindmethod=simple
>        credentials=XXXXXXXX
>        searchbase="o=ec,c=US"
>        type=refreshAndPersist
>        starttls=no
>        tls_reqcert=never
>        interval=00:00:00:10
>        retry="5 5 300 5"
>        timeout=1
>
> mirrormode      true
> overlay         syncprov
> syncprov-checkpoint 100 10
> syncprov-sessionlog 100
>
> So our database is very very low on write/update activity. I think I
> understand that syncprov-checkpoint timed out. I am going to change
> this to
>
> syncprov-checkpoint 10000 9000
> syncprov-sessionlog 10000
>
> As I said our database is very low write, we add users periodically
> and peoples passwords expire that is all the writes/updates that
> happen.
>
> My openldap is 2.4.16 build from source.

Latest stable is 2.4.19.  There have been replication fixes since 2.4.16.

> So important questions :
> 1) Why did two way replication not restart ? If this is the correct
> behavior for my configuration or a bug. Can I configure openldap to
> always start refresh and persist?

Sounds like a bug.  But again, there have been fixes since the release you 
are running.

> 2) Could some entries be out of sync now? If yes
> 2a) Can I use a tool to confirm these systems are in sync? "slapcat &
> diff" (better option)

I'd suggest doing such a diff, yes.

> 2b) if one side is out of sync can i force one side to replicate over the
> other?


man slapd, look at the -c option.

--Quanah

--

Quanah Gibson-Mount
Principal Software Engineer
Zimbra, Inc
--------------------
Zimbra ::  the leader in open source messaging and collaboration