Long-running comet requests are not a problem, they're the system
functioning correctly. A comet request should last up to 120 seconds. If
your request is taking 40 seconds, that means it takes 40 seconds for the
comet actor on the server to send something to the client, which is
perfectly normal. Now, if you're expecting updates more frequently, it's
more likely that the load the backend is causing isn't allowing the comet
actors to actually run, so they don't get a chance to send their responses
back down to the client.
To be sure of what exactly is going on, however, we'd need to see it in
See https://www.assembla.com/spaces/liftweb/wiki/Posting_example_code for
more. Also consider using a profiler like VisualVM or YourKit to get an
inner look at what exactly is taking up CPU time and such.
On Sunday, October 14, 2012 6:41:31 AM UTC-4, Alexander Bertram wrote:
> I develop an application where I use Akka-actors ant Lift. The backend
> causes high cpu load (> 95%). I use comet actors to update the frontend.
> When I start the application, it works as expected. But after a while
> frontend update frequency decreases. At the end it won't be updated at all.
> The backend works as expected.
> After 6 minutes uptime a comet request takes up to 40 seconds (according
> to Chrome Developer Tools).
> Is there something I can do?