> On Sept. 27, 2016, 8:39 a.m., Jonathan Hurley wrote: > > ambari-server/src/main/java/org/apache/ambari/server/upgrade/UpgradeCatalog211.java, > > lines 225-226 > > <https://reviews.apache.org/r/52155/diff/2/?file=1508852#file1508852line225> > > > > This doesn't guarantee ordering of any kind - the first one which comes > > back could have an ID of 9. Since it doesn't match `m_hcsId`, this code > > would try to update it to to a value of 1 and possibly cause duplicates. > > > > You can simply skip this whole section of updating if the > > `resultSet.getLong("id")` is non-null/non-zero. > > > > You can then so something like this: > > `m_hcsId.set(NumberUtils.max(m_hcsId.get(), idKey)` > > wang yaoxin wrote: > Hi Jonathan ,Thanks for your advice, that is highly appreciated. However, > consider the randomness of the value of idKey, my method may avoid this > issue. For instance, if the first-time get value of resultSet.getLong("id") > is 2, and the second-time get value is 1, then basing on your method, the > duplicates error would still occur. Hope we could get further agreement on > this case, thanks.
The problem here is on the 2nd time through, correct? If the first time through it only did half of them (for some reason) then you're left with some rows having an ID and some not. My point is to totally skip rows which have a valid ID - you can sort by ID so you go through all of them first, incrementing m_hceId along the way. This way, you're going to be guaranteed to go through your non-zero IDs first and when you finally hit your first null ID, you're atomic counter is ready to pickup. - Jonathan ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/52155/#review150545 ----------------------------------------------------------- On Sept. 27, 2016, 8:03 a.m., wang yaoxin wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/52155/ > ----------------------------------------------------------- > > (Updated Sept. 27, 2016, 8:03 a.m.) > > > Review request for Ambari, Jonathan Hurley, Oliver Szabo, Robert Nettleton, > and Sebastian Toader. > > > Bugs: AMBARI-18441 > https://issues.apache.org/jira/browse/AMBARI-18441 > > > Repository: ambari > > > Description > ------- > > My first upgrade test failed due to my environment issue. Since in real > production environment this failure would occur again. > > ambari upgrade to 2.1.1 and later version, if the first time failed , excute > ambari-server upgrade again will error duplicate key value violates unique > constraint "pk_hostcomponentstate". > > > Diffs > ----- > > > ambari-server/src/main/java/org/apache/ambari/server/upgrade/UpgradeCatalog211.java > db13612 > > Diff: https://reviews.apache.org/r/52155/diff/ > > > Testing > ------- > > the unit test is UpgradeCatalog211Test.java: Process finished with exit code > 0. > done ! > > > Thanks, > > wang yaoxin > >
