Can I update session variable in a new thread?


#1

Hi,

I have a controller code which does the following

t1= Thread.new {
#some code
session[:executing] = “done”
}

Then in one of the periodic update methods if I access the
session[:executing] it is not updated to “done”. Is there a problem if I
try to update session variable in a different thread?

Regards,
Sudhi


#2

On May 26, 10:20 am, Sudhi K. removed_email_address@domain.invalid
wrote:

session[:executing] it is not updated to “done”. Is there a problem if I
try to update session variable in a different thread?

If this ends up setting the session after rails has closed out the
session (ie pushed it back into the db, or sent the cookie back to the
client) then this won’t work.

Fred


#3

Frederick C. wrote:

On May 26, 10:20�am, Sudhi K. removed_email_address@domain.invalid
wrote:

session[:executing] it is not updated to “done”. Is there a problem if I
try to update session variable in a different thread?

If this ends up setting the session after rails has closed out the
session (ie pushed it back into the db, or sent the cookie back to the
client) then this won’t work.

Fred

How do I know whether rails has closed out and pushed the cookie back to
client? If I add a join before exiting from the action something like
below

def execute
t1 = Thread.new { #code
session[:executing] =“done”;
}

t1.join
end

In the above case the session variable is updated. How would I get this
working if I want the session to be updated sometime later and still not
in that function?

Regards,
Sudhindra


#4

On May 27, 4:36 am, Sudhi K. removed_email_address@domain.invalid
wrote:

Frederick C. wrote:

How do I know whether rails has closed out and pushed the cookie back to
client?

There’s going to be a pretty narrow window after your action method
returns and before rails has got as far as session cleanup. I can’t
think you’d gain anything by waiting for your thread to exit there
rather than in your action.

Fred


#5

Sudhi K. wrote:

Then in one of the periodic update methods if I access the
session[:executing] it is not updated to “done”. Is there a problem if I
try to update session variable in a different thread?

Actually, this is precisely what I was trying to do in another thread
here where Fred graciously helped me too. I’d advise NOT trying to use
the session array Rails gives to you, but instead just telling the
thread the sessionID, and ten doing all the session access yourself
manually. That’s what I’m doing. It’s not TOO hard, although it
inconveniently changes for Rails 2.2 vs. pre-Rails 2.2.

See this:

http://www.ruby-forum.com/topic/184782

And see if it makes any sense to you, or still seems worth it to you
when you see what you must do. :slight_smile: But it does seem to be working for
me, with some basic unit tests.

There’s sadly one more added wrinkle with regard to race conditions when
you start to deal with sessions like this, that can result in the main
action over-writing the session data the thread tries to write, or
vice-versa. I have a hacky way to minimize (although not totally
eliminate that), that is Session-store agnostic. You can totally
eliminate it if you’re willing to hack the particular session store, as
I believe Fred has also done for the AR store, but I think this is
quickly getting beyond what you hoped would be an easy answer.

But it is do-able, although not easy, I don’t think you can get away
with just trying to access the session array itself in a thread that’s
out of the request loop.


#6

But it is do-able, although not easy, I don’t think you can get away
with just trying to access the session array itself in a thread that’s
out of the request loop.

Sorry, this is a topic of interest to me that noone else is ever
interested in talking about.

PPS: REALLY, if you decide you want to set off a thread that will
persist after the response is returned, despite the general pain in the
neck of doing that cause neither ActiveRecord nor MRI 1.8 Ruby are
entirely happy with it (but it’s still possible if you take the right
precautions)…

AND… you need that thread to store state that later requests can
access…

Consider just storing it in the db as an ActiveRecord rather than trying
to put it in a session. All in all easier to deal with, at least if your
thread already was using ActiveRecord anyway (which supplies it’s own
headaches). Worth considering, it’s a trade-off as to what sort of
headaches you want to deal with.


#7

Jonathan R. wrote:

But it is do-able, although not easy, I don’t think you can get away
with just trying to access the session array itself in a thread that’s
out of the request loop.

I’d add it’s do-able ONLY if you are using a server-side session store.
NOT if you are using the cookie store that’s the new Rails default. With
the cookie store there’s simply no possible way to write to the session
outside the request loop. So if you want to write to a session in a
thread, you’d have to wait() on the thread before returning from your
action, which might defeat the purpose you were trying to use the thread
for in the first place.

Also, if you are planning on using ActiveRecord in this thread, you’ve
got to make sure to do a few other things to ensure your AR calls are
thread-safe. Things which change slightly in different Rails versions,
as Rails community has changed it’s consensus on what sorts of threading
are ‘supported’ in what ways.

And ALSO, even aside from AR, I learned the hard way that (at least if
you’re using the standard 1.8 MRI ruby interpreter), if you want to send
a thread off and let it keep going outside the request-loop, you’d
really better set it’s priority to -1, or it’s going to end up
interfering with your response being returned even though you don’t
think it should.

Contrary to popular belief, it IS possible to do concurrent programming
like this in Rails, in all versions. But it’s a big pain in the neck.
Don’t do it unless you really have to (which I think I do in my
situation), try to find another solution if possible.