Updates the UUID package we use from the now neglected "github.com/twinj/uuid" to "github.com/google/uuid"
Calls the uuid.EnableRandPool() setting on initialisation wherever it is used (tests revealed that without this setting, partition key skew is higher by a factor of something between 0.5 - 1% - which is small but did result in a small difference in distribution of data across shards in test when sending data to kinesis)
Changes the kinesis source to generate new UUIDs for partition keys instead of re-using the existing one. This fixes the issue described here
Jira ref: PDP-1431
This PR:
"github.com/twinj/uuid"
to"github.com/google/uuid"
uuid.EnableRandPool()
setting on initialisation wherever it is used (tests revealed that without this setting, partition key skew is higher by a factor of something between 0.5 - 1% - which is small but did result in a small difference in distribution of data across shards in test when sending data to kinesis)