Hi all,
I have searched for the solution of this problem and found none, for
this reason I am writing here.
Basically, what I intend to do is pretty easy, and can be achieved with
several typical solutions easily. However I want to know if there is a
'rails way' to do this. Particularly, a 'rails 3 way', since I went with
Rails 3 for this project.
I have a typical Rails application, in which the special thing is that I
am accessing through SSH to some selected Cisco routers (in order to
read some infos from it). Since it is a public applicacion, it could be
that too many requests come at the same time and the application can
literally 'crash' the router because of too many work.
I want to know if there is some recommended way of setting a limit on
how many people can concurrently request something to the controller.
In order to run the application in production mode I will do it with
Passenger + Apache/nginx.
Best regards, and thanks in advance,
Rafael Fernández López.
Hi,
Since creating a rails app to proxy your SSH connections, you can have
your rails application to handle the frontend (passenger + Apache/
nginx) and show stats or whatever you want.
For the data collection, you can have a background task that connect
via SSH, send commands and store the results in DB (which is read by
the rails app).
With this method, you will be able to manage multiple connection at
the same time. If you also need to send command to reconfigure the
router, you can also create a command queue and execute it with
another background task (or the same) (this command queue could be
filled with a rails app, check resque/redis).
My 2 cent help.
best regards
Christophe
Hi,
Yeah, that could be a solution if that was the case. I didn't explain
enough in depth the problem.
When you perform a request, you provide certain arguments that are
cointained in the request itself, so it's not only a matter of having a
background job that collects information updating the database.
User A requests some kind of action over a selected router, and provides
some arguments (e.g. ping a certain IP provided by the user).
So, what I need is to set some kind of limit of how many requests can be
concurrently running. This could of course be set on the router itself
by limiting the number of SSH sessions, but just wanted to know if there
was some 'rails way' of performing this check.
I think I will do a typical solution like shared memory or some kind of
lock.
Best regards,
Rafael Fernández López.
Hi all,
Finally, the way I fixed this issue has been by creating an initializer
that allocates a hash in which I have each router as a key and the
number of concurrent connections to it. When a connection is going to be
stablished, +1, when the SSH connection ends, -1.
On the controller I check for the value for the desired routers, and if
the case, a warning message of 'overload' is reported on the router that
exceeds that number of maximum connections.
I don't care when the server is restarted, since that is a really short
time window and I also don't care about losing the number of connections
on a restart. It's just a matter of not being able to 'fry' the router
with too many requests at a time.
Thanks for the time and help.
Best regards,
Rafael Fernández López.
