pan-users
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Pan-users] Re: server priority and completeless : EXAMPLE


From: Bernt Hansson
Subject: Re: [Pan-users] Re: server priority and completeless : EXAMPLE
Date: Fri, 27 Mar 2009 11:56:08 +0100
User-agent: slrn/1.0.8 (FreeBSD)

Duncan said the following on 2009-03-27 04:34:
Alan Meyer <address@hidden> posted
address@hidden, excerpted below, on  Thu, 26
Mar 2009 15:46:45 -0700:

I think you just found the problem.

It sure looks like Motzarella dropped the group.

That's possible, but unlikely. Beartooth just started trying motzerella a few days ago, so he just downloaded their list of groups a few days ago. They'd have had to have quit carrying the group in that time, and I don't find that likely.

The more likely case might be that the motzerella server is having problems and is listing headers for messages it doesn't have and losing its own groups.

<rant mode>

I don't know what news server software it's running,

200 news.motzarella.org InterNetNews NNRP server INN 2.5.0 (rc1 version)

INN

but that sounds a lot like some of the problems highwinds-media, the outsourced provider Cox (my ISP) is using, has had.

Highwinds-media is running typhoon

http://www.highwinds-software.com/products/typhoon.php

And other ISPs and providers I've been with over the years, running the same highwinds software, have had similar problems. Apparently, when that software is configured and running perfectly, it has some of the best performance numbers in the business, but it's just terribly difficult to keep it configured just right, and normal changes in usage patterns move even previously properly configured and peak performing servers out of their zone within months, to the point that my experience has been they typically spend perhaps two months working and four months sort of working, cycling maybe twice a year, so a site is lucky to get good service out of them for four months out of the year. The rest of the time they kind of limp along, one bug after another, tweaking this and that and replacing components here and there, until a few months later everything clicks again and they're back at peak performance for another month or two.

Other server software apparently has much lower peak bytes and articles per hour served numbers, meaning on paper more servers are needed to serve the same number of people, but it's easier to maintain at peak performance so there's not quite the same ups and downs. Some of this other software actually runs on commodity x86 hardware, not the "big iron" Sun servers Highwinds software runs on, but it can take fifty or a hundred such servers to handle the job 3-5 Highwinds servers can handle with ease, but only when the Highwinds servers are at peak performance, which as I said my experience as a user suggests they really aren't, more of the time than they are. But there's a reason Google runs reasonably off the shelf commodity hardware... and does so good at it, even if at their size they're changing out several bad servers (out of tens of thousands) a day.

The problem with that is that I hadn't thought motzerella was that big a service, given the cost of big iron and highwinds software and the fact that motzerella is free. So I just don't know.

</rant mode>

Meanwhile, I'm almost getting a hankerin' to sign up and check motzerella out myself, if only just to see what behavior I see here, possibly trying the same messages beartooth is having problems with. If necessary, I could then run a telnet session, and see exactly what the transactions are doing on the wire. Plus then I could see what servers they /are/ running.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]