On 2012-06-07 15:25, Tobias Wunden wrote:
Ruben,
it's usually not recommended to be running the SOLR search indexes off network
(NFS especially) shares, see Google for details.
Tobias
Hi
Sorry for somewhat repeating myself and reviving an old thread.
I have been banging my head against this one for quite a few months now.
I have to choose between restarting Matterhorn on the engage server
every time a new recording is published, or copy the solr indexes from
the core server to the engage player every publish. Until now we have
chosen the former as it is more convenient.
In my mind Matterhorn's service driven architecture should provide a way
for he publishing method to make the engage server rebuild its indexes.
I have not been able to see any indication that the core server in any
way communicates with engage servers solr service to say "here is a new
recording, please update your index"
Digging a bit deeper into the problem with a shared index over nfs makes
the engage servers solr service confused as the indexes has gotten new
file names.
My third attempt have been to make the share mount as 'ro' on engage
player. That borked the entire server, so not a good solution.
So my questions are:
-How is solr supposed to be updated on the engage service if not via a
shared device?
-Is it possible to force the engage solr service to reload the index files?
-How are the other adapters solving a multiserver-setup?
Mark: solr here is embedded solr services and not a separate solr server
(which some of the documentation are referring to)
Id be grateful for any advice!
On Jun 1, 2012, at 3:08 PM, Tobias Wunden wrote:
Hi Frank,
from a thread at the second of May, I read that the best way to
minimize the duplication of data in a cluster is to put the entire
org.opencastproject.storage.dir on a shared NFS drive. Which I did.
only two selected directories should go onto a shared drive, which are the
file.repo.path and the workspace.dir (both from config.properties). All the
other directories *must not* got into shared locations, they are to be kept per
host, otherwise you will experience nodes overwriting each other's data.
Tobias
--
Olav Bringedal
Seksjon for integrasjon og applikasjonsutvikling
IT-Avdelingen UIB
_______________________________________________
Matterhorn-users mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
