Conrad Burger skrev: > Hi > > We are currently experiencing multiple problems with pymsnt. > > We have pymsnt 0.10.2 setup on FreeBSD 5.4 connecting to an ejabberd 0.9.8 > cluster. > > The gateway works well until about 1000 users log on to it. It then starts to > log errors until it comes to a halt. We then have to kill -9 the process and > start it up again. > > No expert here, but I have an idea. (*jumping in the pool*)
I have a hunch that this is the ulimit problem, the limitation of more than 1000 (1064 for FreeBSD I read somewhere) open files for a single shell. Not that I know much about how pymsnt opens files and such but the 1000 number rings a bell for me. Maybe try to run a "ulimit -n 10000" before starting the transport? Remember it has to be run in the shell that starts the transport (it is not system wide). Best regards Stian B. Barmen From [EMAIL PROTECTED] Wed Jan 25 10:30:50 2006 From: [EMAIL PROTECTED] (Norman Rasmussen) Date: Wed Jan 25 10:30:59 2006 Subject: [py-transports] Windows Live Messenger Message-ID: <[EMAIL PROTECTED]> I just got an invite for Windows Live Messenger, and was wondering how this would affect PyMSNt. I assume that PyMSNt uses an old protocol, so it'll just look like I'm using an old client, right? Are there plans to support the new offline messenging of WLM in the future? Should I be looking out for invites to send to you, so that you can look at figuring out the new protocol? PS: (WLM - wasn't that an unknown client flag in tracebacks?) -- - Norman Rasmussen - Email: [EMAIL PROTECTED] - Home page: http://norman.rasmussen.co.za/ From [EMAIL PROTECTED] Wed Jan 25 11:49:32 2006 From: [EMAIL PROTECTED] (James Bunton) Date: Wed Jan 25 11:49:39 2006 Subject: [py-transports] pymsnt multiple problems In-Reply-To: <[EMAIL PROTECTED]> References: <[EMAIL PROTECTED]> <[EMAIL PROTECTED]> Message-ID: <[EMAIL PROTECTED]> On 25/01/2006, at 9:16 PM, Stian B. Barmen wrote: > Conrad Burger skrev: >> Hi >> We are currently experiencing multiple problems with pymsnt. >> We have pymsnt 0.10.2 setup on FreeBSD 5.4 connecting to an ejabberd >> 0.9.8 >> cluster. >> >> The gateway works well until about 1000 users log on to it. It then >> starts to >> log errors until it comes to a halt. We then have to kill -9 the >> process and >> start it up again. >> >> > No expert here, but I have an idea. (*jumping in the pool*) > > I have a hunch that this is the ulimit problem, the limitation of more > than 1000 (1064 for FreeBSD I read somewhere) open files for a single > shell. Not that I know much about how pymsnt opens files and such but > the 1000 number rings a bell for me. Maybe try to run a "ulimit -n > 10000" before starting the transport? Remember it has to be run in the > shell that starts the transport (it is not system wide). > > Best regards > Stian B. Barmen You need to run the transport with a different reactor. By default select() is used, which only supports 1024 connections. You're using BSD, so kqueue would be the best choice, followed by poll if that doesn't work. To set the reactor have a look at the docs on http://msn-transport.jabberstudio.org --- James
