After looking at the cassandra code a little, I believe this is not really an 
issue.
After the upgrade to 1.2, we still see the issue described in this bug I filed:
https://issues.apache.org/jira/browse/CASSANDRA-5264

The "Replicas" is calculated by adding up the effective ownership of all the 
nodes, and chopping off the remainder.  So if your effective ownership is 
299.99%, it appears the code will will report the number of replicas as 2.  
This might become reliably 3 after I complete running repairs after upgrade.
Thanks for you time,-Mike
      From: Alain RODRIGUEZ <arodr...@gmail.com>
 To: user@cassandra.apache.org; Michael Theroux <mthero...@yahoo.com> 
 Sent: Tuesday, June 16, 2015 4:43 PM
 Subject: Re: Nodetool ring and "Replicas" after 1.2 upgrade
   
Hi Michael,
I barely can access internet right now and was not able to check outputs on my 
computer, yet first thing that come to my mind is that since 1.2.x (and vnodes) 
I use rather nodetool status instead. What is the "nodetool status" output ?
Also did you try to specify the keyspace ? Since RF is a per keyspace value, 
maybe this would help.
Other than that, I don't have any idea. I don't remember anything similar, but 
it was a while ago. I have to ask... Why staying so much behind the current 
"stable" / "production ready" version ?
C*heers,
Alain
2015-06-16 14:57 GMT+02:00 Michael Theroux <mthero...@yahoo.com>:



Thanks Jason,
No errors in the log.  Also the nodes do have a consistent schema for the 
keyspace (although this was a problem during the upgrade that we resolved using 
the procedure specified here: 
https://wiki.apache.org/cassandra/FAQ#schema_disagreement).  
-Mike  
      From: Jason Wee <peich...@gmail.com>
 To: user@cassandra.apache.org; Michael Theroux <mthero...@yahoo.com> 
 Sent: Tuesday, June 16, 2015 12:07 AM
 Subject: Re: Nodetool ring and "Replicas" after 1.2 upgrade
   
maybe check the system.log to see if there is any exception and/or error? check 
as well if they are having consistent schema for the keyspace?
hth
jason


On Tue, Jun 16, 2015 at 7:17 AM, Michael Theroux <mthero...@yahoo.com> wrote:

Hello,
We (finally) have just upgraded from Cassandra 1.1 to Cassandra 1.2.19.  
Everything appears to be up and running normally, however, we have noticed 
unusual output from nodetool ring.  There is a new (to us) field "Replicas" in 
the nodetool output, and this field, seemingly at random, is changing from 2 to 
3 and back to 2.
We are using the byte ordered partitioner (we hash our own keys), and have a 
replication factor of 3.  We are also on AWS and utilize the Ec2snitch on a 
single Datacenter.  
Other calls appear to be normal.  "nodetool getEndpoints" returns the proper 
endpoints when querying various keys, nodetool ring and status return that all 
nodes appear healthy.  
Anyone have any hints on what maybe happening, or if this is a problem we 
should be concerned with?
Thanks,-Mike




   



  

Reply via email to