Yeah, HBASE-2670 is bad, but the hang that can happen in 0.20.4 is very bad (and the failed replay of deletes on crash). We need to get a replacement out there. I'm well game for 0.20.6 coming right after a 0.20.5.
St.Ack On Tue, Jun 8, 2010 at 6:11 PM, Ryan Rawson <[email protected]> wrote: > Should we ship 0.20.5 with the fixes as is? The bug is very rare, and > while some people might come across it, it is in 0.20.4 which is known > to be bad. > > We need to get a better 0.20 out there than 0.20.4. What we have is > pretty good, and we can fix the next issue in 0.20.6. > > -ryan > > On Tue, Jun 8, 2010 at 5:42 PM, Todd Lipcon <[email protected]> wrote: >> Unfortunately -1: https://issues.apache.org/jira/browse/HBASE-2670 affects >> branch as well. >> >> I'm working on a fix, will target both branch and trunk, but may be a couple >> days before we can thoroughly test it (this bug is devious) >> >> -Todd >> >> On Mon, Jun 7, 2010 at 1:09 PM, Stack <[email protected]> wrote: >> >>> I've posted an hbase 0.20.5 release candidate 2 here: >>> >>> http://people.apache.org/~stack/hbase-0.20.5-candidate-2/ >>> >>> Should we release this candidate as hbase 0.20.5? >>> >>> It has 18 fixes: http://su.pr/1prW9y >>> >>> Please take it for a test spin if you have a chance. Vote closes >>> Friday, the 11th of June. >>> >>> Yours, >>> Your HBase Homies >>> >> >> >> >> -- >> Todd Lipcon >> Software Engineer, Cloudera >> >
