Some job events may have bad metrics (e.g. totalMappers) #83

Open
sagemintblue opened this Issue Jul 22, 2013 · 1 comment

Comments

Projects
None yet
2 participants
@sagemintblue
Member

sagemintblue commented Jul 22, 2013

Some job events may be including bad job metrics. Once example we've seen is totalMappers, totalReducers: Sometimes values for these metrics will be reported as zero in events towards the end of a job's execution. This may be due to an issue with the PPNL impl not properly collecting this data for some events.

Note that the client UI could be updated to simply avoid overwriting non-zero metric values with zeros, though this may not be the best solution for non-monotonically increasing job metrics (are there any? perhaps this is not an issue?).

@ghost ghost assigned billonahill Jul 22, 2013

@billonahill

This comment has been minimized.

Show comment Hide comment
@billonahill

billonahill Jul 22, 2013

Contributor

We should avoid logic in the client UI that overrides the data given to it. When nodes or tasks fail, the number of completed tasks could decrease, which is a valid scenario.

Contributor

billonahill commented Jul 22, 2013

We should avoid logic in the client UI that overrides the data given to it. When nodes or tasks fail, the number of completed tasks could decrease, which is a valid scenario.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment