cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Safdar Kureishy <>
Subject Re: RandomPartitioner is providing a very skewed distribution of keys across a 5-node Solandra cluster
Date Mon, 25 Jun 2012 06:23:31 GMT
Hi Jake,

Thanks. Yes, I forgot to mention also that I had raised the param from 4 to 5 (to match the # of nodes). Should
I have raised it to 10 or 15 (multiple of 5)? I have added all the
documents that I needed to the index now. It appears the distribution
became more even at a later stage, after indexing 12 million Nutch
documents. The distribution is now 35G / 35G / 56G / 324M / 51G, but there
is still one node that has a small fraction (i.e 324M) of what the other
nodes have. In addition, some nodes also have about double the data as
others (e.g., 56G vs 35G). If you think that increasing param will further improve the distribution, what
would I need to do to enforce that change when the cluster is running, now
that all the data has already been added to the index? And on the flip
side, if the change cannot be made for existing data, what would happen (to
existing + new data) if the setting was changed and the servers were

Lastly, is there another mailing list I should be using for Solandra
questions? I couldn't find one....


On Mon, Jun 25, 2012 at 4:16 AM, Jake Luciani <> wrote:

> Hi Safdar,
> If you want to get better utilization of the cluster raise the
> param in
> -Jake
> On Sun, Jun 24, 2012 at 11:00 AM, Safdar Kureishy <
>> wrote:
>> Hi,
>> I've searched online but was unable to find any leads for the problem
>> below. This mailing list seemed the most appropriate place. Apologies in
>> advance if that isn't the case.
>> I'm running a 5-node Solandra cluster (Solr + Cassandra). I've setup the
>> nodes with tokens *evenly distributed across the token space*, for a
>> 5-node cluster (as evidenced below under the "effective-ownership" column
>> of the "nodetool ring" output). My data is a set of a few million crawled
>> web pages, crawled using Nutch, and also indexed using the "solrindex"
>> command available through Nutch. AFAIK, the key for each document generated
>> from the crawled data is the URL.
>> Based on the "load" values for the nodes below, despite adding about 3
>> million web pages to this index via the HTTP Rest API (e.g.:
>> http://9.9.9.x:8983/solandra/index/update....), some nodes are still
>> "empty". Specifically, nodes and have just a few kilobytes
>> (shown in *bold* below) of the index, while the remaining 3 nodes are
>> consistently getting hammered by all the data. If the RandomPartioner
>> (which is what I'm using for this cluster) is supposed to achieve an even
>> distribution of keys across the token space, why is it that the data below
>> is skewed in this fashion? Literally, no key was yet been hashed to the
>> nodes and below. Could someone possibly shed some light on
>> this absurdity?.
>> [me@hm1 solandra-app]$ bin/nodetool -h hm1 ring
>> Address         DC          Rack        Status State   Load
>>  Effective-Owership  Token
>>                  136112946768375385385349842972707284580
>>       datacenter1 rack1       Up     Normal  7.57 GB
>> 20.00%              0
>>       datacenter1 rack1       Up     Normal  *21.44 KB*
>>  20.00%              34028236692093846346337460743176821145
>>       datacenter1 rack1       Up     Normal  14.99 GB
>>  20.00%              68056473384187692692674921486353642290
>>       datacenter1 rack1       Up     Normal  *50.79 KB*
>>  20.00%              102084710076281539039012382229530463435
>>       datacenter1 rack1       Up     Normal  15.22 GB
>>  20.00%              136112946768375385385349842972707284580
>> Thanks in advance.
>> Regards,
>> Safdar
> --

View raw message