Description
Per a customer reported issue, XDCR needs to behave better on a non-KV node by not repeatedly restarting pipeline and spamming the logs with errors.
In a source cluster is created with one non-KV node, and the source cluster has XDCR Replication activated, the non-KV node will continuously try to create Pipeline and fail because it does not have any vbuckets assigned to it.
PipelineManager 2017-06-20T20:40:00.897-07:00 [INFO] Pipeline cfd2897d2febbd0f24ac8d171ebfa3cd/event_history/event_history is broken, but not yet attended, launch updater
|
PipelineManager 2017-06-20T20:40:00.899-07:00 [INFO] This node is not a KV node, would not act on replication spec cfd2897d2febbd0f24ac8d171ebfa3cd/event_history/event_history's update
|
PipelineManager 2017-06-20T20:40:01.854-07:00 [INFO] Pipeline cfd2897d2febbd0f24ac8d171ebfa3cd/aecbin/aecbin is broken, but not yet attended, launch updater
|
PipelineManager 2017-06-20T20:40:01.855-07:00 [INFO] This node is not a KV node, would not act on replication spec cfd2897d2febbd0f24ac8d171ebfa3cd/aecbin/aecbin's update
|
PipelineManager 2017-06-20T20:40:02.469-07:00 [INFO] Pipeline cfd2897d2febbd0f24ac8d171ebfa3cd/dp_cache/dp_cache is broken, but not yet attended, launch updater
|
PipelineManager 2017-06-20T20:40:02.475-07:00 [INFO] This node is not a KV node, would not act on replication spec cfd2897d2febbd0f24ac8d171ebfa3cd/dp_cache/dp_cache's update
|
PipelineManager 2017-06-20T20:40:03.092-07:00 [INFO] Pipeline cfd2897d2febbd0f24ac8d171ebfa3cd/dp_token/dp_token is broken, but not yet attended, launch updater
|
PipelineManager 2017-06-20T20:40:03.094-07:00 [INFO] This node is not a KV node, would not act on replication spec cfd2897d2febbd0f24ac8d171ebfa3cd/dp_token/dp_token's update
|
PipelineManager 2017-06-20T20:40:03.094-07:00 [INFO] Replication Status = map[cfd2897d2febbd0f24ac8d171ebfa3cd/dp_token/dp_token:name={cfd2897d2febbd0f24ac8d171ebfa3cd/dp_token/dp_token}, status={Pending}, errors={[]}, progress={}
|
cfd2897d2febbd0f24ac8d171ebfa3cd/event_history/event_history:name={cfd2897d2febbd0f24ac8d171ebfa3cd/event_history/event_history}, status={Pending}, errors={[]}, progress={}
|
cfd2897d2febbd0f24ac8d171ebfa3cd/aecbin/aecbin:name={cfd2897d2febbd0f24ac8d171ebfa3cd/aecbin/aecbin}, status={Pending}, errors={[]}, progress={}
|
cfd2897d2febbd0f24ac8d171ebfa3cd/dp_cache/dp_cache:name={cfd2897d2febbd0f24ac8d171ebfa3cd/dp_cache/dp_cache}, status={Pending}, errors={[]}, progress={}
|
]
|
ReplicationManager 2017-06-20T20:40:15.140-07:00 [INFO] Updating status for paused replication cfd2897d2febbd0f24ac8d171ebfa3cd/dp_token/dp_token. kv_vb_map=map[], total_docs=0, docs_processed=0, changes_left=0
|
ReplicationManager 2017-06-20T20:40:15.208-07:00 [INFO] Updating status for paused replication cfd2897d2febbd0f24ac8d171ebfa3cd/event_history/event_history. kv_vb_map=map[], total_docs=0, docs_processed=0, changes_left=0
|
In 4.5.1, the errors bubble up to the UI. In Spock, they do not because of the fix for another issue: MB-25041.
The correct behavior is for XDCR not to detect this situation and handle it gracefully. More design detail may come in the comments.
Attachments
Issue Links
- is caused by
-
MB-23929 Improve replication status updater
- Closed