Anytime I upgrade to new version I export the database using slapcat, and after upgrading re-import the data. If I am not wrong that is the recommended practice (http://www.openldap.org/doc/admin24/maintenance.html ).

Thanks, Mahmudul Hasan System Support University of Lethbridge, AB.

On Sat, Aug 15, 2015 at 12:36 PM, Quanah Gibson-Mount quanah@zimbra.com wrote:

2.4.42 is a bug fix release for significant issues that were in 2.4.41. No new features were added. Primarily for bugs in slapd-mdb and the lmdb library itself. I would upgrade to 2.4.42.

As for your question about different versions there is no issue replicating between 2.4.39 and 2.4.41 or .42

I would note that Howard checked in some potential syncrepl optimizations for slapd-mdb into the master branch this week that may help with your replication delays if you want to test them on top of 2.4.42.

--Quanah

On Aug 15, 2015, at 6:58 AM, Brian Wright brianw@marketo.com wrote:

We have a 2-way master production cluster actively running 2.4.39. We are wanting to upgrade to 2.4.41 (we haven't yet internally certified 42). What is the best practices around upgrading this cluster while keeping at least one of the servers online? Is it replication-safe to have one node on 2.4.39 and one on 2.4.41 for a short period?

If not, I can temporarily comment out the syncrepl statements until both servers are on 2.4.41 and then reenable replication after both are upgraded. We can live with short inconsistency period between the servers during the upgrade of the servers until we can resume replication. Also, is the DB format fully compatible between these two versions or would it be better to reload the DB new from an LDIF backup on 2.4.41? Is there an upgrade doc somewhere that discusses these issues?

Thanks.

--

*Brian Wright* *Sr. UNIX Systems Engineer * 901 Mariners Island Blvd Suite 200 San Mateo, CA 94404 USA *Email *brianw@marketo.com *Phone *+1.650.539.3530 *www.marketo.com http://www.marketo.com/* <Marketo.jpg>

Hi Quanah,

Thanks for the helpful answers. Since the upgrade seems relatively straightforward, I'll work on upgrading to 2.4.42 soon.

On 8/15/15 2:40 PM, Quanah Gibson-Mount wrote:

For migrations between minor or major releases, 2.3->2.4 for example) that would be the correct procedure. For point releases it is almost never necessary. Exceptions being:

1. for BDB based backends, if the BDB major or minor version that

OpenLDAP is linked to is different

2. early releases of slapd-mdb and current slapd-mdb, as there was an

unavoidable db format change. I forget exactly when but I believe somewhere on the 2.4.2x releases.

Outside of that, a full reload is generally pointless, although it can reclaim unused space in the DB created by object deletions.

--Quanah

On Aug 15, 2015, at 2:01 PM, Mahmudul Hasan <cs.mahmud@gmail.com mailto:cs.mahmud@gmail.com> wrote:

...

Anytime I upgrade to new version I export the database using slapcat, and after upgrading re-import the data. If I am not wrong that is the recommended practice (http://www.openldap.org/doc/admin24/maintenance.html).

Thanks, Mahmudul Hasan System Support University of Lethbridge, AB.

On Sat, Aug 15, 2015 at 12:36 PM, Quanah Gibson-Mount <quanah@zimbra.com mailto:quanah@zimbra.com> wrote:

2.4.42 is a bug fix release for significant issues that were in
2.4.41. No new features were added. Primarily for bugs in
slapd-mdb and the lmdb library itself.  I would upgrade to 2.4.42.

As for your question about different versions there is no issue
replicating between 2.4.39 and 2.4.41 or .42

I would note that Howard checked in some potential syncrepl
optimizations for slapd-mdb into the master branch this week that
may help with your replication delays if you want to test them on
top of 2.4.42.

--Quanah


On Aug 15, 2015, at 6:58 AM, Brian Wright <brianw@marketo.com
<mailto:brianw@marketo.com>> wrote:

...

We have a 2-way master production cluster actively running
2.4.39. We are wanting to upgrade to 2.4.41 (we haven't yet
internally certified 42). What is the best practices around
upgrading this cluster while keeping at least one of the servers
online? Is it replication-safe to have one node on 2.4.39 and
one on 2.4.41 for a short period?

If not, I can temporarily comment out the syncrepl statements
until both servers are on 2.4.41 and then reenable replication
after both are upgraded. We can live with short inconsistency
period between the servers during the upgrade of the servers
until we can resume replication. Also, is the DB format fully
compatible between these two versions or would it be better to
reload the DB new from an LDIF backup on 2.4.41? Is there an
upgrade doc somewhere that discusses these issues?

Thanks.

-- 

*Brian Wright*
*Sr. UNIX Systems Engineer *
901 Mariners Island Blvd Suite 200
San Mateo, CA 94404 USA
*Email *brianw@marketo.com <mailto:brianw@marketo.com>
*Phone *+1.650.539.3530 <tel:%2B1.650.539.3530>**
*****www.marketo.com <http://www.marketo.com/>*

	<Marketo.jpg>