cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Pieter Callewaert <pieter.callewa...@be-mobile.be>
Subject RE: forceUserDefinedCompaction in 1.1.0
Date Mon, 02 Jul 2012 13:15:57 GMT
Hi,

While I was typing my mail I had the idea to try with the new directory layout.
It seems you have to change the parameter settings from 1.0 to 1.1
In 1.0:
Param 1: <keyspace name>
Param 2: <datafile name>
In 1.1:
Param 1: <keyspace name>
Param 2: <column family name>/<datafile name>

Don't know if this is a bug or a breaking change ?

Kind regards,
Pieter Callewaert

From: Pieter Callewaert [mailto:pieter.callewaert@be-mobile.be]
Sent: maandag 2 juli 2012 15:10
To: user@cassandra.apache.org
Subject: forceUserDefinedCompaction in 1.1.0

Hi guys,

We have a 6-node 1.0.9 cluster for production and 3-node 1.1.0 cluster for testing the new
version of Cassandra.
In both we insert data in a particular CF with always a TTL of 31 days.
To clean up the files faster  we use the forceUserDefinedCompaction to manually force compaction
on the old sstables which we are sure the data has expired.
In 1.0 this works perfect, but in the 1.1 the command executes without error, but in the log
of the node I see the following:

INFO [CompactionExecutor:546] 2012-07-02 15:05:11,837 CompactionManager.java (line 337) Will
not compact MapData024CTwT/MapData024CTwT-HOS-hc-21150: it is not an active sstable
INFO [CompactionExecutor:546] 2012-07-02 15:05:11,838 CompactionManager.java (line 350) No
file to compact for user defined compaction

I am pretty sure the sstable because it should contain partly expired data en partly data
which is still active.
Does this have to do something with the new directory structure from 1.1 ? Or are the parameters
changed from the function?

Kind regards,
Pieter Callewaert


Mime
View raw message