N1QL query not finding documents

We just upgraded to the GA. I did an insert of 161 new documents and when I did a like query to count them, it only found 135. I wrote a quick search to find the ones that were not in the count thinking there was something wrong in the insert and was surprised to see that the missing documents are in fact in the bucket.

Further research shows that even if I do the N1QL query to expressly request one of the missing documents from the like and they are not found even when search for the meta document id which shows the document in the CLI.

Very puzzled here. Also very worried now on the reliability of N1QL. It’s always the same documents that are missing.

Any thoughts? 6 node cluster.

You need to use AT_PLUS or REQUEST_PLUS scan consistency (see docs). This is by design, so there’s no reliability issue. N1QL indexes are maintained asynchronously.

If you want to search by document id, use the USE KEYS syntax.

Please post for any additional help.