Launchpad itself

  1. Bug #618019
  2. Comment #12

Comment 12 for bug 618019

Revision history for this message
James Henstridge (jamesh) wrote : Re: OOPS may be underrepresenting storm/sql time

I mentioned this on IRC to Robert:

For psycopg2, the Cursor.fetch*() calls are not network IO bound: the network IO occurs during the PQexec() invocation. The fetch*() calls are just involve converting the PGresult to Python values. This is going to be CPU bound, and because it is working with Python objects, it is CPU time holding the GIL.

It doesn't represent additional database round trips.