Do we have kind of native compression in PB? Or not at all? Because if so, it might be an issue.
I run a 0.94 cluster with a mix of 32 and 64 with Snappy enabled. But never tried 0.96 or more on it... 2014-09-15 18:36 GMT-04:00 Andrew Purtell <[email protected]>: > Only where we touch the native Hadoop libraries I think. If you have > specified compression implemented with a Hadoop native library, like > snappy or lzo, and have forgotten to deploy 64 bit native libraries, > and move to this 64 bit environment, you won't be able to open the > affected table(s) until native link issues are resolved. > > > On Mon, Sep 15, 2014 at 2:27 PM, Otis Gospodnetic > <[email protected]> wrote: > > Hi Esteban, > > > > Sorry, I meant to point that out in my original email - yeah, heap sizes, > > Xmx, and such will be different for 32-bit and 64-bit servers, but I was > > wondering if there is anything in HBase that could complain if, say, a > > region written on a 32-bit server moves to a 64-bit server.... or > anything > > else of that sort that is HBase (or Hadoop) specific. > > > > Thanks, > > Otis > > -- > > Monitoring * Alerting * Anomaly Detection * Centralized Log Management > > Solr & Elasticsearch Support * http://sematext.com/ > > > > > > > > On Mon, Sep 15, 2014 at 5:19 PM, Esteban Gutierrez <[email protected] > > > > wrote: > > > >> Well, many HBase internal limits are based on the architecture so that > >> should impact you right away in resource utilization in the RSs, so a > heap > >> or a setting that worked fine for your 32-bit RSs might need to be tuned > >> for a 64-bit environment. Probably might be easier to run a 32-bit JVM > in > >> the 64-bit boxes and phase out the old 32-bit nodes after you test for a > >> while in a couple of the 64-bit nodes new settings. > >> > >> cheers, > >> esteban, > >> > >> > >> -- > >> Cloudera, Inc. > >> > >> > >> On Mon, Sep 15, 2014 at 1:37 PM, Otis Gospodnetic < > >> [email protected]> wrote: > >> > >> > Hi, > >> > > >> > We are thinking of adding some new 64-bit servers to run > RegionServers to > >> > our 32-bit HBase cluster. Anything we should worry about or pay extra > >> > attention to? > >> > > >> > Thanks, > >> > Otis > >> > -- > >> > Monitoring * Alerting * Anomaly Detection * Centralized Log Management > >> > Solr & Elasticsearch Support * http://sematext.com/ > >> > > >> > > > > -- > Best regards, > > - Andy > > Problems worthy of attack prove their worth by hitting back. - Piet > Hein (via Tom White) >
