A simple way you can change the number of vbuckets is by using an 
environment variable. I've done this for a variety of performance tests I 
did sometime last year and it works well providing you follow my caveats.

For example in bash, set the value then start the node or for sanity 
explicitly add the value into the init.d script.

> export COUCHBASE_NUM_VBUCKETS=512

> /etc/init.d/couchbase start


The lowest you can go is a value of 4, to go lower requires changing another 
config variable which I cannot assert won't blow up if changed as I never 
needed to go lower.


*Caveats*:


   - *Changing this variable is not supported by Couchbase*! So only do this on 
non-production systems for your own experimentation.
   - I recommend using a new cluster. Don't change this on a running or 
existing yet dormant cluster.
   - Do this before creating any buckets. Note that existing buckets would lose 
data because some of the existing VB files become out-of-range.
   - It's safest to set this on each nodes before you start the cluster, I've 
never asserted if the cluster manager will propagate this variable to each node 
and at least this allowed me to destroy the cluster and restart from any node.


 

On Monday, 11 May 2015 14:13:51 UTC+1, Vitaly Elyashev wrote:
>
> Hi,
> How I can lower number of vbuckets?
>
> On Monday, October 21, 2013 at 8:53:26 PM UTC+3, Aliaksey Kandratsenka 
> wrote:
>>
>>
>>
>>
>> On Mon, Oct 21, 2013 at 8:33 AM, <[email protected]> wrote:
>>
>>> i will ask again, since couchbase 2.2 was released :) any changes?
>>>
>>
>> As pointed out above expect just relatively minor improvements. People 
>> could run tons of buckets with special tuning. I.e. lowering number of 
>> vbuckets. But there's still pretty significant overhead.
>>
>>  

-- 
You received this message because you are subscribed to the Google Groups 
"Couchbase" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to