Build : 6.5.0-4926
Last known good build : 6.5.0-4924
Steps to reproduce :
1. 2 node cluster with kv+index+query on all nodes running 5.1.0 release
2. Install all 3 sample buckets
3. Run query - select city from `travel-sample` where city is not null; - should return 7341 docs
4. Do an offline upgrade on both the nodes to 6.5.0-4926 build.
5. Post upgrade, observe that the indexes do not have all the docs as before the upgrade.
6. Running query in step#3 returns fewer docs - 19 in my environment. This is data loss post upgrade.
7. Create index - create index idx1 on `travel-sample`(name,city) partition by hash(meta().id) with
- never completes, is stalled.
Performed the same test with different combinations:
1. 5.1.0 -> 6.5.0-4926 = issue is seen
2. 5.1.0 -> 6.5.0-4924 = issue is not seen.
3. 5.0.1 -> 6.5.0-4926 = issue is seen
This means this issue is directly related to the fix that went in 6.5.0-4925 for
|For Gerrit Dashboard: MB-37209|
|119149,4||MB-37209: Set maxVisibleSeqno to highSeqno after off-line upgrade||mad-hatter||kv_engine||Status: MERGED||+2||+1|
|119219,1||Merge branch 'mad-hatter'||master||kv_engine||Status: MERGED||+2||+1|