See Graphite service documentation on using the Graphite service in InfluxDB.
The parsing errors are from malformed data sent to carbon-c-relay that are not valid graphite line protocol data. Carbon-c-relay does not actually check data for correctness, only that it gets an expected time stamp and graphite path. You could use a regex on carbon-c-relay to filter bad data, disable the clients sending them, or just let influx drop them on ingestion.
In the service documentation is a section about templates, which I highly recommend you use for best influx performance. By default, all graphite ingested data will be put in separate, unique, series which will quickly lead to poor influxdb performance.
You might also be interested in InfluxGraph which provides the other part needed to replace Graphite with InfluxDB - Graphite query support using an influxdb data store. It essentially translates Graphite queries to influx queries, with additional graphite-like metric browsing/searching functionality.
It supports the same template functionality as the Graphite service in influxdb to enable use of tags and fields in influxdb schemas via Graphite queries.