There are about four million rows in mlab-oti.raw_ndt.scamper1 dataset that show they took more than 30 minutes to complete. Some show a surprising difference of about 330,000 seconds (~91 hours) between cycle stop and cycle start times (raw.CycleStop.stop_time - raw.CycleStart.start_time).
What is not clear is that even though the early versions of traceroute-caller did not support a timeout mechanism when invoking scamper, one would expect scamper itself would time out if a traceroute took longer than a reasonable amount of time.
Fortunately, these strange rows are in the period from 2019-09-26 to 2019-12-12 and make up only about 0.1% of all rows.
Filing this issue to have a record for later reference and investigation.
There are about four million rows in
mlab-oti.raw_ndt.scamper1
dataset that show they took more than 30 minutes to complete. Some show a surprising difference of about 330,000 seconds (~91 hours) between cycle stop and cycle start times (raw.CycleStop.stop_time - raw.CycleStart.start_time).What is not clear is that even though the early versions of
traceroute-caller
did not support a timeout mechanism when invokingscamper
, one would expectscamper
itself would time out if a traceroute took longer than a reasonable amount of time.Fortunately, these strange rows are in the period from 2019-09-26 to 2019-12-12 and make up only about 0.1% of all rows.
Filing this issue to have a record for later reference and investigation.