Best Operation practice in production

classic Classic list List threaded Threaded
3 messages Options
kcheng.mvp kcheng.mvp
Reply | Threaded
Open this post in threaded view
|

Best Operation practice in production

This post has NOT been accepted by the mailing list yet.
I am going to build cache/grid compute platform using ignite. I have one question about the best operation practice.


Suppose the cache system is in production already, and right now I need to implement another new cache.
per my understanding about ignite I have to shutdown all the server node, and restart it, as the new cache introduce a new pojo, I have to put it into ${node_server_home}/libs. but this practice is not acceptable. as our other service is using this cache.


I had thought P2P class loading is the Silver Bullet , but after some round clarifications
https://issues.apache.org/jira/browse/IGNITE-1190 and find p2p class loading is just for compute.


from the document https://apacheignite.readme.io/docs/zero-deployment

Ignite will check if class is available on local classpath (i.e. if it was loaded at system startup)

Ignite load all the class in the folder libs at system startup, so there is no change to load a class dynamically.


So I think if Ignite can load a class on demand at runtime, then we can also implement a new cache without server node restart.

Does Ignite support load a class on demand at runtime?






vkulichenko vkulichenko
Reply | Threaded
Open this post in threaded view
|

Re: Best Operation practice in production

kcheng.mvp wrote
Suppose the cache system is in production already, and right now I need to implement another new cache.
per my understanding about ignite I have to shutdown all the server node, and restart it, as the new cache introduce a new pojo, I have to put it into ${node_server_home}/libs. but this practice is not acceptable. as our other service is using this cache.
Hi,

Your understanding is correct and right now there is no proper way to achieve this. We're currently working on changes in marshalling which will remove the requirement to have data classes on server nodes. This will allow to start a client with new classes, start a new cache and put instances of these classes into the cache without deploying them on data nodes. In my understanding this will solve your issue, correct me if I'm wrong.

Here is the ticket that you can track: https://issues.apache.org/jira/browse/IGNITE-950

-Val
kcheng.mvp kcheng.mvp
Reply | Threaded
Open this post in threaded view
|

Re: Best Operation practice in production

Thank you very much, once this feature is available, I would like to be the first use to test it and use it.