cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rahul Neelakantan <ra...@rahul.be>
Subject Re: Repair taking long time
Date Mon, 29 Sep 2014 19:19:40 GMT
What is the recommendation on the number of tokens value? I am asking because of the issue
with sequential repairs on token range after token range.

Rahul Neelakantan

> On Sep 29, 2014, at 2:29 PM, Robert Coli <rcoli@eventbrite.com> wrote:
> 
>> On Fri, Sep 26, 2014 at 9:52 AM, Gene Robichaux <Gene.Robichaux@match.com>
wrote:
>> I am fairly new to Cassandra. We have a 9 node cluster, 5 in one DC and 4 in another.
>> 
>>  
>> 
>> Running a repair on a large column family seems to be moving much slower than I expect.
>> 
> 
> Unfortunately, as others have mentioned, the slowness/broken-ness of repair is a long
running (groan!) issue and therefore currently expected. 
> 
> At this time, I do not recommend upgrading to 2.1 in production to attempt to fix it.
I am also broadly skeptical that it as fixed in 2.1 as all that.
> 
> Once can increase gc_grace_seconds to 34 days [1] and repair once a month, which should
help make repair slightly more tractable.
> 
> For now you should probably evaluate which of your column families you *absolutely must*
repair (because you do DELETE like operations in them, etc.) and only repair those.
> 
> As an aside, you "just lose" with vnodes and clusters of the size. I presume you plan
to grow over appx 9 nodes per DC, in which case you probably do want vnodes enabled.
> 
> One note :
>>  Looking at nodetool compaction stats it indicates the Validation phase is running
that the total bytes is 4.5T (4505336278756).
> 
> This is the uncompressed size, I'm betting your actual on disk size is closer to 2T?
Even though 2.0 has improved performance for nodes with lots of data, 2T per node is still
relatively "fat" for a Cassandra node.
> 
> 
> =Rob
> [1] https://issues.apache.org/jira/browse/CASSANDRA-5850

Mime
View raw message