Replication Strategy for Jackrabbit

[dave_gough <ba...@googlemail.com>]

Hi,

Can anyone help regarding the best way of migrating and syncing from one
"live" repository to a "fail-over" repository on a totally different server?
Here are some background facts:

1) We are using the
org.apache.jackrabbit.core.persistence.bundle.MySqlPersistenceManager.
2) Blobs are stored externally on disk (a SAN) and are up to 1 TB
3) Indexes are about 300MB
4) We typically have about 100MB of multimedia content ingested into the
repository every hour. 

We were initially thinking of using MySQL replication for the database and
rsync for the indexes and blobs (after an initial transfer of blobs by
external disk) to start things off and also as an ongoing means of
replicating the repsotiry to a fail-over system, but we have concerns about
the consistency between the data in MySQL and the indexes/blobs on the file
system.

Some question we have are:

How fault-tolerant is the repository if there are missing blobs referenced
by the database. 

Is there a cleanup that occurs if a blob is not found or is a warning
issued. I am aware of the "consistencyCheck/Fix" attributes in the
repository.xml config and the GarbageCollector which we run every day at
02h00. I presume these all affect any inconsistencies between the database
and FileSystem in some way.

We are not afraid of losing some data, so if we have to fail over and some
data is missing that is OK as we can just re-ingest any missing content, but
we are concerned about corrupting the repository through inconsistent
database/file-system and indexes.

We cannot use Jackrabbit clustering as the FileSystems are not shared as one
of the servers is an off-site fail-over server.

Does this seem a workable solution, or is there a better way of doing this,

regards

Dave Gough

-- 
View this message in context: http://n4.nabble.com/Replication-Strategy-for-Jackrabbit-tp1475531p1475531.html
Sent from the Jackrabbit - Users mailing list archive at Nabble.com.