You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@kylin.apache.org by ShaoFeng Shi <sh...@apache.org> on 2017/12/07 09:40:09 UTC

Re: kylin streaming cube merge strategy?

Hi, Could you share the merge settings, and the logs when a new segment be
built? If not ready for auto merge, Kylin will print a message "Not ready
for merge on cube". Please provide detail log for troubleshooting.

2017-12-07 14:38 GMT+08:00 446463844@qq.com <44...@qq.com>:

> Hi:
> I try to use kylin streaming cube for building cube from kafka.
> And when I design the cube I set cube merge options on advanced Settings
> panel.
> when I build cube for times .I found the merge strategy not valid
> Can somebody help?
>
>
> 446463844@qq.com
>



-- 
Best regards,

Shaofeng Shi 史少锋

Re: Re: kylin streaming cube merge strategy?

Posted by "446463844@qq.com" <44...@qq.com>.
 Hi:
I use streaming cube for days .and auto build per 10 mintues
in monitor page I find no auto merge job is   triggered.
And  mine merge options be setting :
0.5 hour
1 hour
2 hour
4 hour
8 hour
1 day
7 day
Is there all  most of segment not reached the merge section?



446463844@qq.com
 
From: ShaoFeng Shi
Date: 2017-12-08 16:08
To: dev
Subject: Re: kylin streaming cube merge strategy?
Correct; Its merge is not by offset, but by start/end time of the segments;
A timestamp field on message is required.
 
2017-12-08 15:51 GMT+08:00 funky <fu...@gmail.com>:
 
> thx,I try to understand it .Streaming cube build success and named with
> offset rather than timestamp means that it will be merged with internal
> timestamp (data structure  must contains a 'timestamp' field in kafka).is
> that rigtht ?
>
> --
> Sent from: http://apache-kylin.74782.x6.nabble.com/
>
 
 
 
-- 
Best regards,
 
Shaofeng Shi 史少锋

Re: kylin streaming cube merge strategy?

Posted by ShaoFeng Shi <sh...@apache.org>.
Correct; Its merge is not by offset, but by start/end time of the segments;
A timestamp field on message is required.

2017-12-08 15:51 GMT+08:00 funky <fu...@gmail.com>:

> thx,I try to understand it .Streaming cube build success and named with
> offset rather than timestamp means that it will be merged with internal
> timestamp (data structure  must contains a 'timestamp' field in kafka).is
> that rigtht ?
>
> --
> Sent from: http://apache-kylin.74782.x6.nabble.com/
>



-- 
Best regards,

Shaofeng Shi 史少锋

Re: kylin streaming cube merge strategy?

Posted by funky <fu...@gmail.com>.
thx,I try to understand it .Streaming cube build success and named with
offset rather than timestamp means that it will be merged with internal
timestamp (data structure  must contains a 'timestamp' field in kafka).is
that rigtht ?

--
Sent from: http://apache-kylin.74782.x6.nabble.com/

Re: kylin streaming cube merge strategy?

Posted by ShaoFeng Shi <sh...@apache.org>.
Hi,

The merge will check the start/end time of adjacent cube segments; If their
time has reached one of the merging thresholds, it will trigger a merge job.

For example, the min. and max. timestamp of 3 segments are:

seg1: 1:00 to 1:05
seg2: 1:05 to 1:20
seg3: 1:18 to 1:35

Assume you have defined 30 minutes as a threshold. Then when seg3 is built,
Kylin found seg1 to seg3 has reached this threshold, then it will merge
them. After the merge, the new segment is:
seg': 1:00 to 1:35

If you didn't see a merge be triggered, some possible situations:
1) there is hold among the segments; (the merging segments need be
continuous on Kafka)
2) the start/end time of each segment already exceeds some merging
threshold, but not reach a bigger threshold.

Please double check, or share more information for the investigation.

2017-12-08 13:33 GMT+08:00 funky <fu...@gmail.com>:

> er. I can manual merge the cube now,So can somebody tell me the streaming
> cube merge strategy automatic ?
>
> --
> Sent from: http://apache-kylin.74782.x6.nabble.com/
>



-- 
Best regards,

Shaofeng Shi 史少锋

Re: kylin streaming cube merge strategy?

Posted by funky <fu...@gmail.com>.
er. I can manual merge the cube now,So can somebody tell me the streaming
cube merge strategy automatic ?

--
Sent from: http://apache-kylin.74782.x6.nabble.com/

Re: kylin streaming cube merge strategy?

Posted by funky <fu...@gmail.com>.
yes I found "Ngt ready for merge on cube streaming_cube n cube " in kylin.log
I wonder that when I maunl merge cube segment in kylin Web UI and options
message that 
Find no segments to merge by data range 1512518400000-1512604800000 for
CUBE[mate=streaming_version_cube]

the merge strategy when I design this cube
Auto Merge Thresholds
0.5hour
1hour
2hour
4hour
8hour
1day
7day
28days

--
Sent from: http://apache-kylin.74782.x6.nabble.com/