[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
News...
- To: outages-list@eskimo.com
- Subject: News...
- From: Robert Dinse <nanook@eskimo.com>
- Date: Fri, 28 Jan 2000 02:03:21 -0800 (PST)
- Resent-Date: Fri, 28 Jan 2000 02:03:26 -0800
- Resent-From: outages-list@eskimo.com
- Resent-Message-ID: <"z1FUT1.0.C37.ibMau"@mx1>
- Resent-Sender: outages-list-request@eskimo.com
I apologize for the authentication problems on the news server...
As many of you know we have had an ongoing stability problem with the news
server since upgrading to INN version 2.2.2 from version 1.7.2.
I had posted to several relevant newsgroups relating to this problem in
the past and not received any useful suggestions but tonight someone e-mailed
me with some useful info. They had experienced the same problem and chased it
down to a bug in the libraries and provided a fix.
I applied the fix and it did in fact fix the stability problem. However;
I started with virgin source since I had hacked the heck out of it trying to
fix the problem and none of those things I had done had corrected it. One
problem though, our authentication method here is non-standard and that
requires a small hack to the nnrpd program (nntp read/post daemon) that I
neglected to re-apply, and that broke authentication for everyone.
That is fixed, innd now appears to be stable, and this should be pretty
much all the futzing I'll do with this server until we are ready to turn up the
new CPU. Changes in the Redhat 6.1 environment is making it difficult to
compile some necessary code, in particular code relating to authentication, on
the new system. As soon as we can get those problems resolved we'll be placing
that on-line.
A problem with the current box is that it gets it's total CPU power from
four moderate speed CPU's but innd being a single process that consumes the
bulk of the CPU can't take advantage of more than one CPU and thus can get at
most 25% of the available CPU power, it needs more. So we are working on
replacing it with a machine that has approximately the same CPU performance but
in a single CPU so that innd can use what it needs. It also eliminates some
spin-lock problems that cause infrequent crashes (was responsible for todays).
But between now and that time things should at least be relatively stable
now.