Details
-
Bug
-
Resolution: Duplicate
-
Critical
-
6.5.0
-
None
-
Untriaged
-
Unknown
Description
Summary:
All documents on a bucket are lost when a node joins the cluster.
To Reproduce:
- Have a cluster of a single node that has data service.
- Load beer-sample bucket and make sure item count matches the expected.
- Add another node to the cluster with any service (e.g. Analytics).
Even before the new node is rebalanced in, the item count on beer-sample goes to zero.
Logs from both nodes are attached. The following log appears in the memcached file when the second node joins the cluster:
2019-04-17T03:04:30.774715+03:00 INFO 41: [ 127.0.0.1:60128 - 127.0.0.1:11999 (<ud>@ns_server</ud>) ] Updated cluster configuration for bucket [beer-sample]. New revision: 19
|
2019-04-17T03:04:30.774779+03:00 INFO Pushing new cluster config for bucket:[beer-sample] revision:[19]
|
2019-04-17T03:04:30.775766+03:00 INFO 46: Client 127.0.0.1:60243 authenticated as <ud>@ns_server</ud>
|
2019-04-17T03:04:30.775986+03:00 INFO 46: HELO [regular] [ 127.0.0.1:60243 - 127.0.0.1:11999 (<ud>@ns_server</ud>) ]
|
2019-04-17T03:04:30.932795+03:00 INFO (beer-sample) VBucket::setState: transitioning vb:1023 from:active to:dead
|
2019-04-17T03:04:30.932846+03:00 INFO (beer-sample) Deletion of vb:1023 was completed.
|
2019-04-17T03:04:30.933439+03:00 INFO (beer-sample) VBucket::setState: transitioning vb:1022 from:active to:dead
|
2019-04-17T03:04:30.933468+03:00 INFO (beer-sample) Deletion of vb:1022 was completed.
|
2019-04-17T03:04:30.933626+03:00 INFO (beer-sample) VBucket::setState: transitioning vb:1021 from:active to:dead
|
2019-04-17T03:04:30.933652+03:00 INFO (beer-sample) Deletion of vb:1021 was completed.
|
...
|
Attachments
Issue Links
- duplicates
-
MB-33828 Adding a node into a cluster wipes all the documents in default bucket
- Closed