When I restart the rancher server, all registered hosts need to register again, it's unacceptable.

[Jason Huang]

It's too bad, if we managed 100 server, when restart the rancher server, I need to restart all agents of the 100 server again. It's really unacceptable.

[qsys]

About everyone realizes that this is not acceptable. High availability is something that's worked on. More info will be published once in beta. I had pretty similar concerns:
https://groups.google.com/forum/#!topic/rancherio/lDr2KCvR88g
Remember that the project is still pretty young and very actively being developed. There are still some very important (and critical) features, such as high availability, in development. It's just a matter of time.

qsys


Op woensdag 13 mei 2015 11:33:52 UTC+2 schreef Jason Huang:

[Darren Shepherd]

That is unacceptable, you're right.  The current behavior is that you only need to reregister the agent if the server is reinstalled not restarted.  If the server is restarted agents will automatically reconnect.  In fact if the server is upgraded and a newer agent is required, it will auto upgrade to the right version of the agent.  We try very hard to design this thing to work at scale.

Can you explain what you're doing that leads you to believe a server restart requires a manual restart of the agent.  Maybe we have miscommunicated something or you have found a bug.

Thanks,

Darren

On May 13, 2015, at 2:33 AM, Jason Huang <jason...@gmail.com> wrote:

It's too bad, if we managed 100 server, when restart the rancher server, I need to restart all agents of the 100 server again. It's really unacceptable.

--
You received this message because you are subscribed to the Google Groups "Rancher" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rancherio+...@googlegroups.com.
To post to this group, send email to ranc...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/rancherio/046ee363-9229-4366-b38f-45d5d2bd0e37%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Jason Huang]

First of all, I had tested it again, the be behavior is what you had said.

I used the following command you gave to start rancher.

docker run -d -p 8080:8080 rancher/server

I also used the command to start the rancher server for the second time. It start another container, so all data had lost, I need to register the agents again. It's my mistake.

But it's easy to make user lack of experience a litter confuse. It's better to give some hints. :)

I start the stopped container again, it keeps the agents' info.

Next, From my experience for developing management system, it's not good design for managed large scale servers like current method. The following methods maybe better.

1. Provide an auto scan methods,  give the IP sets or IP ranges to scan, and added the agents automatically. 

2. Every agent know the server IP, if the server had not changed, when server start, agents may notify the server to add them again.

3. Giving the IP sets or IP ranges and ssh authentication, server should using ssh method to install the agents and make the hosts(VMs) become the manged hosts(VMs)

4. Like Bosh and Cloud Foundry， we may using the other deployment tool to provision the manged hosts(VM), the provisioned hosts(VM) automatically add them to the managed hosts. It's a little similar as the method 3.

Think about the HA reason and integration with the different system, it's better to provide the option to separate the ranch server and the database and give the user choice to select RDBMS in the future.

The following features maybe needed in the production environment:

1. provide the router to forwarding the request from outside to the app automatically.

2. start the app by not specify the host (assign the host automatically by policies)

3. provide fail-over functions to migrate the unavailable app to other hosts.

... 

Wish the rancher server to be an awesome system. :)

在 2015年5月13日星期三 UTC+8下午9:11:55，Darren S写道：