[ATLAS-3989] Behavior change in Atlas API to get Atlas Server object during metadata replication

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Component/s: atlas-core
• Assignee: Ashutosh Mestry
• Reporter: Ashutosh Mestry
• Created: 09/Oct/20 22:23
• Updated: 12/Oct/20 05:39
• Resolved: 12/Oct/20 05:39

Description

There is a change being noticed in the Atlas server API used during atlas metadata replication in hive.
Earlier the API: '/api/atlas/admin/server/cl1' where 'cl1' is the cluster name, used to work.
Now, it doesn't work any more. This Api is used during replication to get the changeMarker info.
Now the API works with some thing called server name which is 'default' by default.
Also, now the prerequisite is to at least have one export or import API call in order to get this 'default' initialized.
The change-marker is obtained prior to invoking export and hence the all policy is bound to fail unless we mandate calling Export or Import API outside hive replication once to get this object initialized, even if hive uses 'default' in the call.

 

Root cause

We switched from using

atlas.cluster.name

to

atlas.metadata.namespace