HA design assumes shared storage. QJM can be used as such, so as other things. If that was your question. --Konst
On Thu, Sep 27, 2012 at 2:11 AM, Andrew Purtell <apurt...@apache.org> wrote: > I don't follow. For example the QJM and HA NameNode configuration are > designed together to eliminate the SPOF in HDFS, within HDFS. I don't see > how they make sense separately. > > On Thursday, September 27, 2012, Konstantin Shvachko wrote: > >> The SPOF is in HDFS. This project is about shared storage >> implementation, that could be replaced by NFS or BookKeeper or >> something else. >> Suppose failure happened because NFS failed. Where do you go? NetApp. >> Same if QJ failed. You go to the creators, which will be somewhat >> closer than NetApp, because it is still Hadoop. >> >> --Konst >> >> On Thu, Sep 27, 2012 at 12:50 AM, Stack <st...@duboce.net <javascript:;>> >> wrote: >> > On Wed, Sep 26, 2012 at 4:21 PM, Konstantin Shvachko >> > <shv.had...@gmail.com <javascript:;>> wrote: >> >> Don't understand your argument. Else where? >> > >> > You suggest users should download HDFS and then go to another project >> > (or subproject) -- i.e. 'elsewhere' -- to get a fundamental, a fix for >> > the SPOF. IMO, the SPOF-fix belongs in HDFS core. >> > >> > St.Ack >> > > > -- > Best regards, > > - Andy > > Problems worthy of attack prove their worth by hitting back. - Piet Hein > (via Tom White)
