Our customer has a major problem that the size of index volumes become
very large in our customer's production envrironment.
Now the size is 500GB, whichi is 38% of the size of message store
although we estimated it as 25% before.

Our customer wants to reduce the size of the index.

I think one reason for increasing index is that we use Lucene Analyzer
for Japanese language instead of StandardAnalyzer, and I assume
another major reason is that index is actually not being deleted
(at least right away) when we delete messages in the mailbox.

So I ask Zimbra Support for this issue and they replied:

> Our long-term POP customers' mailboxes did show
> that the index was not especially large, and the files in the index
> directory were all touched fairly recently (within the last month), so
> over time the files are being cleaned up. I will monitor my test
> account to see if the mailboxd process takes care of the index as part
> of its table maintenance or purge operations.
> A reindex operation will force the index directory to reduce after
> messages are deleted, but it appears that an automatic system is at
> work as well. I will check on this automatic process and let you know
> what I find.

I am not sure and want to know what is this "automatic process" but
no replies from them for 2 weeks and it seems they are not intrested in
this topic any more...

I searched the source code of ZCS NE 5.0.16 for IndexWriter.optimize()
which seems to reduce(or merge) index segments , but I can't find any
source code executing it.
We are not running zmmaintaintables.

Does anyone know the automatic system for deleting index??