Ram, Yes, each column family in the 4 c/f table has only one column qualifier.
Will try addColumn(byte[] fam, byte[] qual) and test how that performs. Don't have another version of hbase to test across releases, but will see if I can manage this. In test case to scan only rowkeys, are you saying we still need to use addColumnFamily to limit scan to 1 c/f? Here is the code for that test, should addColumnFamily (or addColumn??) be used here, or it will read all column families? return new Scan(startInclusive, endExclusive) .setFilter(new FilterList(FilterList.Operator.MUST_PASS_ALL, new KeyOnlyFilter(true), new FirstKeyOnlyFilter()) ) .setCaching(caching) .setCacheBlocks(false); Thanks. On Mon, Aug 21, 2017 at 1:15 PM, Partha <[email protected]> wrote: > Will send across table statement and the test code. Pls let me know if you > find anything from your test given the inputs so far. Note that column > family has only 1 qualifier with json payload value of size 15KB. The > column families use fastdiff encoding and gzip compression. > > Added user@ to this reply. > > Thanks > > On Aug 21, 2017 12:47 PM, "ramkrishna vasudevan" < > [email protected]> wrote: > >> Ya . Just send replies to user@. Generally using replyAll would have >> done that. >> >> Regarding the test case - I will try to create a simple test case and see >> what is happening. Though I tried out one it seemed things are working as >> expected. >> >> Looks like something obvious is missing here. Can you send across the >> test case or the test program including how the tables are created? >> >> Regards >> Ram >> >> On Mon, Aug 21, 2017 at 10:13 PM, Partha <[email protected]> wrote: >> >>> Not intentional :-) Just haven't figured out how to post to public >>> thread, just send replies to [email protected]? >>> >>> Thought this was a simple test case, to read 1MM rows on each table >>> (limiting 2nd table to 1 cf). Any suggestion to make it simpler? >>> >>> Thanks >>> >>> >>> On Aug 21, 2017 12:38 PM, "ramkrishna vasudevan" < >>> [email protected]> wrote: >>> >>> Thanks for the info. I just noticed that you have not included user@. >>> Was it intentional? People generally don't like side tracking a >>> conversation. >>> How ever I am happy to help you with this issue. >>> >>> Have you tried creating some simple test cases and checked out how it >>> performs? >>> >>> Regards >>> Ram >>> >>> On Mon, Aug 21, 2017 at 10:02 PM, Partha <[email protected]> wrote: >>> >>>> 1.1.2.2.6.0.3-8 >>>> >>>> On Aug 21, 2017 12:40 AM, "ramkrishna vasudevan" < >>>> [email protected]> wrote: >>>> >>>>> Which version of HBase do you use ? >>>>> >>>>> On Sun, Aug 20, 2017 at 4:34 AM, Partha <[email protected]> >>>>> wrote: >>>>> >>>>>> Hi Ramkrishna, >>>>>> >>>>>> Correct. The scan on 2nd table targets a single column family (using >>>>>> addFamily), the first table has only one c/f so not using addFamily >>>>>> there. >>>>>> The column family in each case has 1 qualifier and json value of >>>>>> comparable >>>>>> size. >>>>>> >>>>>> I did not check the cache hit ratio yet on either scan, but I did do >>>>>> a major compaction on both tables and the tests after didn't show any >>>>>> change in test results. >>>>>> >>>>> >>>>> >>> >>> >>
