"clean" shutdown and startup using service wrapper?


(David Taylor) #1

Hi,

I was looking through the service wrapper's init script and realised the "stop" action uses SIGTERM. Does this result in a "clean" shutdown? Is it the same as using the API to shutdown a node?

What if I simultaneously used the init script to stop all nodes? Does that behaviour differ greatly from using the API to shutdown all nodes?

Basically, using the init script feels more Linux-like for managing services... but is it recommended?

Cheers,
David.


(Shay Banon) #2

Yes, a SIGTERM causes a clean shutdown. There is no difference between that and a single node shutdown API call.

The big difference is the shutdown API can actually shutdown a whole cluster in a more optimized manner, without causing ripples as nodes shutdown. The shutdown is executed on the master, which then tells all the other nodes to shutodwn, without treating them as failed nodes, and then, the master shuts down itself.

So, the recommended way to shutdown a cluster is using the shutdown API.

-shay.banon
On Thursday, April 21, 2011 at 2:39 AM, David Taylor wrote:

Hi,

I was looking through the service wrapper's init script and realised the "stop" action uses SIGTERM. Does this result in a "clean" shutdown? Is it the same as using the API to shutdown a node?

What if I simultaneously used the init script to stop all nodes? Does that behaviour differ greatly from using the API to shutdown all nodes?

Basically, using the init script feels more Linux-like for managing services... but is it recommended?

Cheers,
David.


(system) #3