----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/53398/#review154621 -----------------------------------------------------------
Ship it! Ship It! - Jason Huynh On Nov. 2, 2016, 7:26 p.m., Dan Smith wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/53398/ > ----------------------------------------------------------- > > (Updated Nov. 2, 2016, 7:26 p.m.) > > > Review request for geode, Jason Huynh and nabarun nag. > > > Repository: geode > > > Description > ------- > > This is a fix for pretty specific race condition > 1) T1 does a put and gets to the point of calling setIndexBufferTime, > but hasn't updated the indexes > 2) T2 starts a query and finds the entry in the index even though the > value no longer matches the query > 3) T1 finishes the put > 4) T2 checks to see if should revaluate the entry, but decides not to > because based on the SAFE_QUERY_TIME value the entry changed before the > query started. > > By moving the update to SAFE_QUERY_TIME down, if the an entry > doesn't need reevaluation based on the SAFE_QUERY_TIME, we know the > index was updated before the query started. > > There is currently an updateInProgress flag to handle the issue of the > query executing before the SAFE_QUERY_TIME is updated. If the entry is > updated, but not the index, the updateInProgress flag will be set. > > > Diffs > ----- > > geode-core/src/main/java/org/apache/geode/internal/cache/LocalRegion.java > 360c6a9f444b41a99bec1a896e660bd506d793ff > > Diff: https://reviews.apache.org/r/53398/diff/ > > > Testing > ------- > > > Thanks, > > Dan Smith > >
