Is there an option with a pure Java fallback if the native codec isn't available? I mean something reasonable, not bzip2.
> On Dec 19, 2017, at 6:16 PM, Dave Latham <[email protected]> wrote: > > What about LZ4 instead? Most benchmarks I've seen show it ahead of Snappy. > >> On Tue, Dec 19, 2017 at 5:55 PM, Mike Drob <[email protected]> wrote: >> >> Can you file a JIRA for some kind of magical default snappy-if-available? >> >>> On Tue, Dec 19, 2017 at 7:38 PM, Stack <[email protected]> wrote: >>> >>>> On Tue, Dec 19, 2017 at 4:22 PM, Stack <[email protected]> wrote: >>>> >>>> Thanks for jumping in JMS. Ok on the by-table. >>>> >>>> SNAPPY license seems fine. We'd enable it as default when you create a >>>> table? Let me play w/ it. >>>> >>>> >>> Oh. I forgot what happens if the native lib is not available, how cluster >>> goes down. >>> >>> Caused by: java.lang.RuntimeException: native snappy library not >> available: >>> this version of libhadoop was built without snappy support. >>> >>> I think we should skip out on enabling this (but recommend folks run this >>> way...) >>> >>> Thanks JMS, >>> S >>> >>> >>> >>>> Anything else from your experience that we should change JMS? >>>> >>>> Thanks sir, >>>> S >>>> >>>> >>>> On Tue, Dec 19, 2017 at 1:47 PM, Jean-Marc Spaggiari < >>>> [email protected]> wrote: >>>> >>>>> Can we get all tables by default Snappy compressed? I think because of >>> the >>>>> license we can not, right? Just asking, in case there is an option for >>>>> that... Also +1 on balancing by table... >>>>> >>>>> 2017-12-18 17:34 GMT-05:00 Stack <[email protected]>: >>>>> >>>>>> (I thought I'd already posted a DISCUSSION on defaults for 2.0.0 but >>>>> can't >>>>>> find it...) >>>>>> >>>>>> Dear All: >>>>>> >>>>>> I'm trying to get some eyeballs/thoughts on changes you'd like seen >> in >>>>>> hbase defaults for hbase-2.0.0. We have a an ISSUE and some good >>>>> discussion >>>>>> already up at HBASE-19148. >>>>>> >>>>>> A good case is being made for enabling balancing by table as >> default. >>>>>> >>>>>> Guanghao Zhang has already put in place more sensible retry/timeout >>>>>> numbers. >>>>>> >>>>>> Anything else we should change? Shout here or up on the issue. >>>>>> >>>>>> Thanks, >>>>>> S >>>>>> >>>>> >>>> >>>> >>> >>
