cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jack Krupansky <jack.krupan...@gmail.com>
Subject Re: distributing load across cluster
Date Wed, 10 Feb 2016 21:00:27 GMT
That's for one node. You can look at the writes for each node.

I'm actually not sure if the partition key count includes memtables in
addition to sstables. A nodetool flush will assure that any memtable data
gets flushed to sstables.

-- Jack Krupansky

On Wed, Feb 10, 2016 at 3:30 PM, Ted Yu <yuzhihong@gmail.com> wrote:

> Here is output from cfstats:
>
> http://pastebin.com/W4FVd4RW
>
> The keyspace was created as described in
> https://github.com/cloudius-systems/osv/wiki/Benchmarking-Cassandra-and-other-NoSQL-databases-with-YCSB
>
> Data was loaded by using ycsb.
>
> Cheers
>
>
>
> On Wed, Feb 10, 2016 at 12:26 PM, Jack Krupansky <jack.krupansky@gmail.com
> > wrote:
>
>> Sorry, I didn't realize you were still living in the stone age with DSE -
>> and Cassandra 2.1. Chnage "table" to "cf" (column family.)
>>
>> -- Jack Krupansky
>>
>> On Wed, Feb 10, 2016 at 3:23 PM, Ted Yu <yuzhihong@gmail.com> wrote:
>>
>>> I don't see tablestats sub-command:
>>>
>>> http://pastebin.com/XwwCAqh4
>>>
>>> This is DSE 4.8.4
>>>
>>> Cheers
>>>
>>> On Wed, Feb 10, 2016 at 12:05 PM, Jack Krupansky <
>>> jack.krupansky@gmail.com> wrote:
>>>
>>>> What do your partition and cluster keys look like?
>>>>
>>>> Check a nodetool tablestats to see number of partition keys on the
>>>> nodes. Also check nodetool tablehistograms to see if you have a lot of
>>>> too-wide rows due to the balance of data between the partition key and
>>>> clustering columns.
>>>>
>>>> -- Jack Krupansky
>>>>
>>>> On Wed, Feb 10, 2016 at 2:54 PM, Ted Yu <yuzhihong@gmail.com> wrote:
>>>>
>>>>> Hi,
>>>>> I am following this guide on a 5 node cluster:
>>>>>
>>>>> https://github.com/cloudius-systems/osv/wiki/Benchmarking-Cassandra-and-other-NoSQL-databases-with-YCSB
>>>>>
>>>>> I am using ycsb-0.5.0
>>>>>
>>>>> I found that some node receives above average writes, leading to disk
>>>>> full condition.
>>>>>
>>>>> I want to get some suggestion on how the load can be better
>>>>> distributed.
>>>>>
>>>>> Thanks
>>>>>
>>>>
>>>>
>>>
>>
>

Mime
View raw message