Re: Cassandra JMX without RMI?

2023-07-19 Thread Maxim Muzafarov
Hello,

I haven't found any good documentation to refer to, but you can have a
look at the following:
https://github.com/apache/cassandra-sidecar
https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=95652224

The sidecar is deployed alongside the Cassandra node and exposes
metrics/commands via HTTP/REST.

This is what it look like:
https://cwiki.apache.org/confluence/download/attachments/247827919/SBW%20OSS%20SBR.jpg?version=1=1681166031000=v2

On Wed, 19 Jul 2023 at 19:00, Angelo Polo  wrote:
>
> Thanks Max, that's a good option to know about.
>
> Unfortunately I have a fairly strict requirement for an agentless 
> architecture. It's possible to have some values set in cassandra-env for 
> example, but I can't say "download and install this jar here" (Jalokia agent 
> mode) or "run this code from here" (Jalokia proxy mode - Jalokia would still 
> be using JMX over RMI I assume, but the proxy shields my application from 
> having to make RMI calls directly).
>
> Is JMX over RMI the only out-of-the-box way to communicate remotely with 
> Cassandra's MBean server?
>
> Best,
> Angelo
>
> On Tue, Jul 18, 2023 at 11:49 PM Max Campos  wrote:
>>
>> We use Jolokia.  It runs inside of Cassandra and then gives you an HTTP REST 
>> interface to interact with JMX.  Works great for us!
>>
>> https://jolokia.org/
>>
>> - Max
>>
>> On Jul 18, 2023, at 8:11 pm, Angelo Polo  wrote:
>>
>> Is it possible to connect to the Cassandra JMX server without using the RMI 
>> protocol?
>>
>> I have found for example, this open source project 
>> https://github.com/willemsrb/simple-jmx that claims to provide JMX without 
>> RMI, but it seems to need to be set up on the server side too, and of course 
>> Cassandra's JMX server isn't based on this project.
>>
>> So is there a client-side-only possibility to use something other than RMI 
>> when retrieving Cassandra JMX?
>>
>> Thanks,
>> Angelo
>>
>>


Re: Cassandra JMX without RMI?

2023-07-19 Thread Angelo Polo
Thanks Max, that's a good option to know about.

Unfortunately I have a fairly strict requirement for an agentless
architecture. It's possible to have some values set in cassandra-env for
example, but I can't say "download and install this jar here" (Jalokia
agent mode) or "run this code from here" (Jalokia proxy mode - Jalokia
would still be using JMX over RMI I assume, but the proxy shields my
application from having to make RMI calls directly).

Is JMX over RMI the only out-of-the-box way to communicate remotely with
Cassandra's MBean server?

Best,
Angelo

On Tue, Jul 18, 2023 at 11:49 PM Max Campos 
wrote:

> We use Jolokia.  It runs inside of Cassandra and then gives you an HTTP
> REST interface to interact with JMX.  Works great for us!
>
> https://jolokia.org/
>
> - Max
>
> On Jul 18, 2023, at 8:11 pm, Angelo Polo  wrote:
>
> Is it possible to connect to the Cassandra JMX server without using the
> RMI protocol?
>
> I have found for example, this open source project
> https://github.com/willemsrb/simple-jmx that claims to provide JMX
> without RMI, but it seems to need to be set up on the server side too, and
> of course Cassandra's JMX server isn't based on this project.
>
> So is there a client-side-only possibility to use something other than RMI
> when retrieving Cassandra JMX?
>
> Thanks,
> Angelo
>
>
>


LocalQuorum requiring 3 replicas with RF=3 when a node is joining

2023-07-19 Thread Luciano Greiner
Hello.

This company I am working for has small production cluster with the
following setup:

2 DCs
3 nodes each
RF = 3 (all keyspaces)
num_tokens = 4
repairs are up to date

We are in the process of adding more nodes to the cluster, although
while testing our procedures on a staging cluster (same setup), we
constantly get this error while the new node is in JOINING or LEAVING
stage:

"Server timeout during read query at consistency LocalQuorum (2
replica(s) responded over 3 required)"

Trying to understand why would Cassandra require 3 replicas when the
node is joining the cluster, when the quorum should be 2. We have no
such error when one of the three nodes is down as expected.

I have been bouncing around this error for a while now, so any hints
or thoughts would be much appreciated.

Thank you

Luciano