incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From aaron morton <aa...@thelastpickle.com>
Subject Re: Fixing / Rebuilding Secondary Index
Date Wed, 14 Nov 2012 21:39:52 GMT
> But for about 12 hours the secondary index has completed only 3%. 
How much data do you have?

> I use Cassandra 1.0.7 and use CL of ONE for my read requests. 
There is your problem. 
If you use CL ONE for any time type of query you can get inconsistent results. Hinted Handoff
and Read Repair may eventually repair the inconsistency depending on the settings and nature
of the failures. The simple thing to do though is run nodetool repair. 

If you use CL ONE for an indexed query (with no start key) though we only query at the co-ordinator.


Use a higher CL and or run repair. 

Cheers

-----------------
Aaron Morton
Freelance Cassandra Developer
New Zealand

@aaronmorton
http://www.thelastpickle.com

On 15/11/2012, at 8:50 AM, Eran Chinthaka Withana <eran.chinthaka@gmail.com> wrote:

> Hi,
> 
> I have a secondary index for a column family which is connected to a keyspace that spans
over three data centers. I observed that the index is not complete on one of the data centers.
Reason for that conclusion is, I tried to retrieve an object, using the secondary index, in
DC1 and it was a success. Then I used a node in DC2 to retrieve the object the same way but
it returned nothing. 
> 
> Can someone please help me to understand why this is happening and how I can fix this?
Yesterday, through #cassandra IRC, I learnt that one can initiate a secondary index re-build
by calling a refresh (and copying a new sstable) and I did just that. But for about 12 hours
the secondary index has completed only 3%. 
> 
> I use Cassandra 1.0.7 and use CL of ONE for my read requests. 
> 
> Finally, since I want to fix this issue faster, I am thinking of either
> copying full copy of data from DC1 into each and every node in DC2 and start a cleanup/repair,
or
> drop the secondary index and re-create it using "update column family" in CLI
> What do you think about those approaches? 
> 
> Thanks,
> Eran Chinthaka Withana


Mime
View raw message