Hi Ajay Bhullar
This bug fix reverses the fix done for MB-32087. For current fix, it would be good to cover below scenarios:
1. Regression tests covering all three consistency options (Any, Request Plus and Query consistency).
2. Without any rollback or bucket flush: It would be good to run repeated scans with all three consistency options and fast snapshots generation by indexer - for example:
a. MOI storage with simple index on single small field
b. MOI array index on an array that does not exist in the document and MOI storage. Ex: CREATE INDEX `idx1` ON `bucket-1`(all array1)
And, array1 is missing in the document; since array1 is missing, there is not much work done on storage ensuring fast snapshots.
3. Bucket flush: A request_plus scan right after a bucket flush will return results instead of 0. Please note, this point is not a validation of the fix but really just an old behavior we are reverting to. The tracking bug for this issue is MB-32087.
Sharath Sulochana: Please update when the Sync Gateway load tests have run successfully for multiple iterations with the fix.
Hi Ajay Bhullar
This bug fix reverses the fix done for MB-32087. For current fix, it would be good to cover below scenarios:
1. Regression tests covering all three consistency options (Any, Request Plus and Query consistency).
2. Without any rollback or bucket flush: It would be good to run repeated scans with all three consistency options and fast snapshots generation by indexer - for example:
a. MOI storage with simple index on single small field
b. MOI array index on an array that does not exist in the document and MOI storage. Ex: CREATE INDEX `idx1` ON `bucket-1`(all array1)
And, array1 is missing in the document; since array1 is missing, there is not much work done on storage ensuring fast snapshots.
3. Bucket flush: A request_plus scan right after a bucket flush will return results instead of 0. Please note, this point is not a validation of the fix but really just an old behavior we are reverting to. The tracking bug for this issue is MB-32087.
Sharath Sulochana: Please update when the Sync Gateway load tests have run successfully for multiple iterations with the fix.