Forum OpenACS Development: Re: Site node cache memory problems

Collapse
Posted by Jose Agustin Lopez Bueno on
Hi Malte!

The problems with site-nodes are not solutionated
in cluster. We have 4 nodes in the cluster.

Every time we create new communities we must to
load all the site-nodes in cache. All the instances in cluster must refresh his own data and the time for this
process can be greather than 1 minute. Sometimes
all the cluster nodes are blocked and the performance
decrease.

select count(*) from site_nodes;
132929
The select waste 10 seconds in postgres.
The reorder of memory tcl arrays waste up to 1 minute.
The update local cache of every node take a duration
then greather than 1 minute X 4 nodes >= 4 minutes.

For the usability of cluster structure we must to redesign
the policy of cache refresh in the procs in acs-tcl:
site_node::update_cache

Best regards,
Agustin