Hi Marcel,

I did not yet have  thorough look at it. From the Cassandra codebase I see only 
two JVM shutdown hooks being registered, so it seems feasible. On the other 
hand, if it would be that simple and many people ask for it, why is there no 
code example on how to do it? But at least we should analyze the issue 
thoroughly first.

Regards, Ivo

From: [email protected] 
[mailto:[email protected]] On Behalf Of Marcel Offermans
Sent: woensdag 3 augustus 2011 10:42
To: [email protected]
Cc: Dion Meijer; Arthur Meijer
Subject: Re: [Amdatu-developers] Cassandra shutdown

Hello Ivo,

On 3 Aug 2011, at 10:23 , Ivo Ladage-van Doorn wrote:

The Cassandra daemon was not designed to be stopped in ways other than killing 
the JVM. I did already run into that issue (see CassandraDaemonService), but 
without changing Cassandra it just doesn't seem to be possible to support 
runtime updates of the Cassandra daemon bundle. The Cassandra guru's (Jonathan 
Ellis and Ran Tavory) confirm that Cassandra was not designed to facilitate 
shutdown in other ways then killing the JVM.

Did you already analyze if it's feasible to change Cassandra to support proper 
shutdown?

Is this is big issue for provisioning?

Yes, provisioning is pretty useless if we have to shutdown the whole JVM every 
time there is a change in Cassandra.

Our fallback scenario is to fork a new JVM from within our Cassandra bundle and 
manage the life cycle of that separate process to match that of the launching 
bundle. However, it does not feel right to start forking JVMs just because 
there are problems shutting down a single bundle. I don't believe there are 
fundamental reasons why Cassandra cannot shutdown properly, it just was not 
important for their use cases so far.

I would start with doing the analysis, and trying to mobilize other interested 
parties on the Cassandra mailing list to see if something can be done about 
this.

Greetings, Marcel

_______________________________________________
Amdatu-developers mailing list
[email protected]
http://lists.amdatu.org/mailman/listinfo/amdatu-developers

Reply via email to