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

Re: How to debug single object syncrepl failure



On 13/02/15 18:46, Dieter Klünter wrote:
Am Fri, 13 Feb 2015 09:42:22 +1100
schrieb Geoff Crompton <geoffc@trinity.unimelb.edu.au>:

Hi,

I have a particular object in my LDAP database that is failing to
replicate (using syncrepl between two slapd's running 2.4.31-1+nmu2
on Debian Wheezy), despite other objects succeeding to replicate. I'm
not using a 'filter' configuration in my olcSyncrepl configuration
that might exclude this particular object, and I've checked that the
binddn I'm using has permission to see this object all the attributes
of the object that isn't replicating.

[...]

Any tips on how to further debug this?

access rules?
corrupted file system?
corrupted database?

Dieter


I've checked access rules on the producer, and the account the consumer is replicating with has access to the object in question. I've confirmed the object doesn't exist on the consumer by reviewing the "slapcat" output, it's not access rules preventing me seeing the object.

I hope I've avoided a corrupted file system or corrupted database, as I"m seeing the same problem across multiple consumer VMs. At the same time some older VMs (running Debian Squeeze, and slapd 2.4.23-7.3) are not having this problem, which I think rules out corrupted filesystem or database on the producer.

This information supports Quannah's observation that the problem is probably the syncrepl bugs in 2.4.31.

I'm going to upgrade my consumer to the 2.4.40 backport that is now available in Debian Wheezy, and see if the problem goes away (and probably upgrade my producer likewise in my next maintenance window).

--
Geoff Crompton, System Administrator
T: +61 (0)3 9348 7138
Trinity College | University of Melbourne | Royal Parade, Parkville | Victoria 3052, Australia
www.trinity.unimelb.edu.au