Hi Giovanni: On Mon, 21 Jul 2008, Binda Giovanni (bing) wrote: > I run into the following problem: webcoll (running in the scheduler) > doesn't update the cache. I had to run a one time task that force the > cache update (webcoll -f). > > I have the following questions: > > How and when is webcoll updating the cache?
You can run webcoll with verbose 3 level to see some hints: $ sudo -u apache /opt/cds-invenio/bin/webcoll -u admin -v 3 2008-08-06 12:38:40 --> Task #12 submitted. $ sudo -u apache /opt/cds-invenio/bin/webcoll 12 2008-08-06 12:38:45 --> Task #12 started. 2008-08-06 12:38:45 --> Database timestamp is 2008-07-11 18:12:56. 2008-08-06 12:38:45 --> Collection cache timestamp is 2008-08-06 12:26:39. 2008-08-06 12:38:45 --> Collection cache is up to date, no need to run. 2008-08-06 12:38:45 --> Task #12 finished. [DONE] webcoll recreates the cache when the database update timestamp is found to be close to the cache update timestamp within a tolerance of 20 seconds in either direction. > Are the timestamp taken from the database server or from the > application server? The database timestamp is a max of table update times for several search-related tables such as 'collection', 'portalbox' etc. The collection cache timestamp is written by webcoll when the collection cache was last updated and is located on the application server at /opt/cds-invenio/var/cache/collections/last_updated. > Do I need a clock synchronisation (by ntpd for example)? webcoll's cache timestamp comparison facility has a tolerance of 20 seconds. If the clocks on the DB and APP server are further apart, then webcoll may indeed not see the DB changes as being new. The clock synchro will prevent problems in this case. The full-scale 'ntpd' daemon should not be needed, a nighly 'ntpdate' call via cron on both machines should do the job as well. Best regards -- Tibor Simko ** CERN Document Server ** <http://cds.cern.ch/>
