May Zhou created KYLIN-3643:
Summary: Derived column from windowSpec not working in where
Issue Type: Bug
I have reproduced this issue using the sample dataset of kylin, and I've
opened a JIRA to track this issue.
Sent from: http://apache-kylin.74782.x6.nabble.com/
Could you please confirm if filtering of rows (WHERE clause) is done
in CoProcessor side?
Is there any APIs/logging to get physical plan of query? It will help
us in optimising the cube.
On Mon, Oct 22, 2018 at 8:58 PM Shrikant Bang
> Thanks ShaoFeng for response.
You may post your query related log here, there should be some query log that
indicated whether the filter is push down or not, at least in the returned
response stats, there's some log show how many rows are filtered in the
At 2018-10-23 16:58:36, "Sachin Aggarwal" wrote:
I am observing some queries are running for longer time and getting
failed. Here is exception in kylin logs for "BadQueryDetector".
Could someone please help me to understand about Bad Queries and
meaning of below stack trace :
2018-10-23 09:36:40,647 DEBUG
I hope this may help you.
??: "Sachin Aggarwal";
: 2018??10??23??(??) 7:25
: join column derived vs normal
Hubert STEFANI created KYLIN-3644:
Summary: NumberFormatExcetion on null values when building cube
while creating cube what are recommendations across join column between
fact and dimension.
should we mark it as derived or normal while selecting it from dimension
what factors play a role in making this decision.
Thanks & Regards
Thank you Ma for reply.
I have one of dimension of size 1.2GB with records ~ 13 millions and its a
UHC. I am trying with higher size of 'kylin.dictionary.forest-trie-max-mb'
to 2GB and marking 'kylin.dictionary.reusable' true as dimensions are
Are there any other setting which could help
Thank you Ma for response.
I didn't find any logs about applying filters in co-processors. Also I
don't see logs for logical/physical query plans in kylin.log. Am I missing
I would like to put logger statements in code and build specific module(s)
as required for tracing latencies --
The log indicated that your query needs data from a snapshot lookup table(your
query columns should contain some derived columns defined in the cube), but the
snapshot is very large, so the query is very slow. You may check the snapshot
size in the lookup table's snapshot tab in
Hi Sachin Aggarwal
When a lookup table is stored as snapshot, all columns in that table will be
taken as derived dimension by default during cube design.
Compare to normal dimension, derived dimension does not participate in cube
calculation directly. Instead, it is represented by its FK
I tried debugging code to understand query execution (on Kylin v2.5 hbase
1.x) . As I understood query is formed and executed using Calcite APIs.
Could you please help me to get answer of below?
1. With my understanding default implementation of filter operation
Which Kylin version you used? By default, most of the filters will be pushed
down to coprocessor, you may grep "Total filtered row:" in your log to see how
many rows are filtered from coprocessor side.
If you want to change code to add some log there, you can add log at method:
rongchuan.jin created KYLIN-3645:
Summary: Kylin does not clean table metadata when drop
Mail list logo