Hi Gabriele,
I never worked on Cyrus 2.5, so I can't talk from first-hand experience:
On Wed, Nov 25, 2020, at 9:11 AM, Gabriele Bulfon via Info wrote:
When these are very big (90,000 - 130,000 msgs around 10-12GB of data) this reindexing takes very long (30-50 minutes for a big mailbox).I thought incremental would mean also a changing mailbox would be reindexed incrementally, while I think squatter is rebuilding the index from scratch.
I looked at the 2.5 squatter code and it indeed should skip already indexed messages. When you run squatter in verbose mode, it should tell you the number of indexed messages. Does that match the total number of messages in the INBOX?
How can I optimize this?
I guess this is not the answer you are looking for: if you can, upgrade to Cyrus version 3. We completely reworked search there. If not, I hope someone who actively uses 2.5 joins the discussion.
If you want to debug yourself, you could start poking your Cyrus installation directory for directory names starting with "metadir-squat-". These contain the internal index files that keep track of indexed message UIDs per mailbox.
Also, is there any option to run multiple squatter threads in parallel?
AFAIK this isn't an option in 2.5. It definitely is not possible in 3.
Cheers,
Robert
