Hi,

I plan to use Hibernate Search, hopefully with automatic "transparent" index updates (automatically on session commit, without explicit coding).
I just wonder how it handles errors (e.g. on the file system, database connection, etc):

1) Does it issue some event/notification, which I can catch (so I can send email to admin, issue some GUI message, etc)?

2) Can it be configured to re-try (e.g. tell it "if you run into an error, try again every 10 minutes, up to 5 attempts, before you give up").

3) Do the replies above change if I configure changes to be asynchrnous?
Do the replies above change if I configure Hibernate Search to use JMS?

Thanks :)

Hi, welcome

yes, all operations related to event styled indexing can have a custom evenhandler registered:
http://docs.jboss.org/hibernate/search/ ... e/#d0e2226
This is only relevant to indexing operations, stuff like you mention to "open a database connection" is normal hibernate business: so it's hibernate which might throw runtime exceptions to your application. The additional custom eventhandler in search is neede because some work is possibly performed in other threads than your application.


No. Why? all operations to the index are put in the "todo" queue upon a succesfull commit to the database, it's a transaction sync. It's unlikely but possible that operations fail after this queue, for this case the eventhandler mentioned above will be notified and you can handle it ad you wish. The most common failure case would be "disk crashed" or "disk is full", so a retry operation doesn't make sense before the admin does something. In this cases you usually perform a full reindexing the get the index in synch with the db state - we always consider the db as the reliable source of information.


No, same strategies.

using JMS you can make sure that the backend queue is eventually fully processed, it's up to you to code the receiver so you could configure JMS to not consider a message processed if you can't invoke the Search backends without received errors.