I created sg_allDocs_x2 index with expression from sg_allDocs_x1, and query with sg_allDocs_x2
index works just fine. I can not see any pattern for missed entries. Among the missed data there are with xattrs._sync.sequence = 180 and xattrs._sync.sequence = 20012, where max xattrs._sync.sequence ~26000.
So dropping and recreation of sg_* indexes should solve the current problem.
But how can I prevent this problem in the future? In web interface this index marked as green and it is not in updating phase, so at some point I got corrupted index, but no indication that I have the problem. Are any hypothesis why this happens and how can I prevent this?
I use couchbase 6.6.0 from rpm on centos 7.9 if it is important.