SurveyUSA polled the San Diego mayoral race on the eve of the election, showing Kevin Faulconer with a narrow 47%-46% lead. Faulconer won comfortably. SurveyUSA blamed voters for not voting when they said they would. A pollster isn't supposed to report what the voter tells them. A pollster is supposed to report what's actually going to happen. You don't let a voter tell you whether they'll vote. You poll people you believe will vote. California, fortunately, provides information on how many of the last five primary and general elections a voter has voted in and which ones of those they submitted a ballot.
If I were conducting a poll for this primary, I'd call voters who voted in either the June 2010 or June 2012 primaries. The only exceptions I'd make would be those under 20 who were too young to vote in those elections or new registrants. Even then I wouldn't mind under polling these groups because they are less likely to vote. My method might underestimate the electoral pool by 2-3%. Not doing that has caused SurveyUSA to overestimate by about 30%.