Closed gothub closed 2 years ago
The 'last_harvest_date' maintained by metadig-engine for each node harvest, used to be set to the systemmetadata modified time. When the next harvest would happen, the last pid would be re-harvested unnecessarily.
This problem was fixed by setting the last_harvest_date
(in the node_harvest
table to one millisecond after the sysmeta modified time, so only newer pids will be picked up on the next harvest.
Under some situations, pids are being reharvested:
from metadig-scheduler pod log:
This pid should only have been submitted once.