Hi, Nice to meet you, Jackrabbit community :) We have startup performance issues with an embedded Jackrabbit 1.5: the production cluster that holds about ten thousand documents takes over 20 minutes to start up. First off, do you have any performance tips that we may not have tried yet - specifically related to startup time? (Kindly note that we cannot upgrade to version 2, our hands are tied.)
To make things a little more challenging, there's clustering: We have 4 nodes in production and we have previously encountered synchronization issues between these. Documents added by 1 instance would not show up in the other's searches. One stability measure was to delete each node's file-system index before starting up. With a few hundred documents, regenerating the index even seemed faster than verifying it! Now that there are thousands of documents, we consider NOT deleting the index anymore, to help speed up the startup phase. But that would potentially re-introduce cluster sync problems. So we're thinking along the following lines and would need your input to refine this: If we can't prevent the sync problems, then let's at least detect them. Maybe all it needs is periodically comparing the revision ID tables and alerting our support team if nodes do not have the same index revision? Then we would restart the servers with an empty index to fix the problem. Do you think this method would detect all sync failures? Can you think of another approach? We'd love to have your thoughts on this! Have you experienced long startup times? Have you had node synchronization problems? How did you solve these? Best regards, Patrick
