On Thu, Aug 26, 2010 at 8:30 AM, Josh M. Patten <[email protected]>wrote:

> Even at 50 users the current RLS implementation has problems when
> changing/adding any user attribute. This encompasses everything from setting
> a call forward to changing an email address to adding a new user. Each time
> one of these actions occurs the RLS server completely tears down the active
> RLS list in memory and reloads it from scratch (very inefficient, very
> not-scalable). While this is happening the RLS server "pauses" any new BLF
> events which can cause the park monitors to appear locked because the signal
> to terminate the BLF ringing is never sent.
>
> Hoa, one of our programmers here at Brazos County, recently took on the
> challenge of rewriting the offending code in order to make it more efficient
> and scalable and I think he's done it. I am running the RLS fix he wrote on
> our production system (currently at 270 registered endpoints) and have yet
> to see any of the behaviors that the old RLS implementation exhibited. I can
> now make changes, add users, change user settings, etc. without locking up
> the RLS server (with the current implementation once I got to about 200 or
> so users the RLS server would lock up every time I made a change or added a
> user). I have this fix available as a 64 bit RPM for 4.2.x. If anyone is
> desperate for it I can email it to you. The patch will be put in the tracker
> shortly.
>
> -
>

Thanks Josh.  What build is your is your production system?  It would be
worth trying the your patch.  Users at this site use web-ui and chat
extensively so this might explain the increased park issues.  Ping me off
list and I'll test it I'll my lab first.
_______________________________________________
sipx-users mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-users/

Reply via email to