ignite cluster management

[wangsan <wq...@gmail.com>]

I am using ignite to manage my cluster.when node join I save an item in my
NodeCache, when node left I delete an item in my NodeCache with ignite event
listener. 
Use custom NodeCache but not ignite cluster topology,The reason is I want to
keep node status persistence when node left for some restart operations.
I have some problems
1. How to guarantee client node with same consistent(or other properties) do
not join the cluster twice like server node works.
2. My nodes topology will have server 100+ client 2000+. how to make the
cluster keep high stability.

tks



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: ignite cluster management

[Ilya Kasnacheev <il...@gmail.com>]

Hello!

I think you can have a singleton service that periodically checks the
attributes() for all nodes. If anyone is amiss from previous runs, try to
re-start it.

Be sure to use 'exclude neighbours' cache affinity setting when having
multiple server nodes on same machine to avoid data loss.

Regards,
-- 
Ilya Kasnacheev


пн, 3 сент. 2018 г. в 17:53, wangsan <wq...@gmail.com>:

>
> 1. Note that you will have problem writing to caches when topology changes.
> Why not just query ignite.cluster(), expose node traits as
> node.attributes()?
>
> >> Yes, I use attributes for node join. But I want to know the stopped
> >> nodes(when node left) . Then I can do some operations such as restart or
> >> alarm or start on anther machine.
>
> 2. Having 100 servers should not be a problem. It is recommended to have
> long-lived clients in this case, as opposed to creating and taking down
> them
> all the time.
>
> >> In a machine, It will many nodes, When the machine crash. I hope the
> >> cluster can work well.
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>

Re: ignite cluster management

[wangsan <wq...@gmail.com>]

1. Note that you will have problem writing to caches when topology changes.
Why not just query ignite.cluster(), expose node traits as
node.attributes()?

>> Yes, I use attributes for node join. But I want to know the stopped
>> nodes(when node left) . Then I can do some operations such as restart or
>> alarm or start on anther machine. 

2. Having 100 servers should not be a problem. It is recommended to have
long-lived clients in this case, as opposed to creating and taking down them
all the time.

>> In a machine, It will many nodes, When the machine crash. I hope the
>> cluster can work well.



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: ignite cluster management

[Ilya Kasnacheev <il...@gmail.com>]

Hello!

1. Note that you will have problem writing to caches when topology changes.
Why not just query ignite.cluster(), expose node traits as
node.attributes()?

2. Having 100 servers should not be a problem. It is recommended to have
long-lived clients in this case, as opposed to creating and taking down
them all the time.

Regards,
-- 
Ilya Kasnacheev


вт, 28 авг. 2018 г. в 5:39, wangsan <wq...@gmail.com>:

> I am using ignite to manage my cluster.when node join I save an item in my
> NodeCache, when node left I delete an item in my NodeCache with ignite
> event
> listener.
> Use custom NodeCache but not ignite cluster topology,The reason is I want
> to
> keep node status persistence when node left for some restart operations.
> I have some problems
> 1. How to guarantee client node with same consistent(or other properties)
> do
> not join the cluster twice like server node works.
> 2. My nodes topology will have server 100+ client 2000+. how to make the
> cluster keep high stability.
>
> tks
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>