> Would it be possible for DNews to exhibit some more intelligent behaviour
> when it runs out of disk space? It's clearly able to delete old articles
> to make room for new ones because it seems to do this all the time, yet
> when one of the disks making up the spool gets full, it decides to stop
> sucking new news, even though there's plenty of space available on the
> other disks. Shouldn't it still be able to make room for itself, and/or
> use the other disks? As it is, it stores headers but not bodies for a
> while, resulting in groups full of articles that can't be retrieved (and
> which DNews won't resuck because it thinks it's already got them - this
> needs improving too), then after doing this for a while, it finally
> realises that the disk is full and stops everything, usually during our
> overnight binaries suck which is the only time we have spare bandwidth
> and nobody is around to fix it. By the time we can fix it, we have to
> wait for the next overnight download to catch up.
>
> It seems that the only way to avoid this problem is to leave considerable
> slack between the amount of space that DNews is told to use on each disk,
> and the actual usable capacity of the disk. This means we have to run
> with about 1-2GB of the spool unused, which could store a lot of articles
> and is a huge waste. It was to reduce this wastage that I started
> trying to optimise its disk usage, which is when these problems started -
> up to then DNews has been performing well and we've been very pleased
> with it. Could this behaviour be improved? Alternatively, if anybody
> can offer any advice on how to always run on close to 100% disk
> utilisation (we currently have to settle for about 90%) then I'd
> appreciate it. Of course, the best solution would be to install bigger
> disks, but that's currently not an option.
>
> Btw, we're running DNews 5.5b5 on Slackware Linux 7.1. I note that
> there's a newer version available but am reluctant to install it
> without seeing a list of changes - is there one available?
Hi, try setting in dnews.conf
spool_free true
it will not fix the above problem you really have to make the settings low
enough so it will not run out as the indexes are written by the time the
message
body is written it's too late to 'efficiently' undo the damage. Also
running the
disk completely full will tend to lower performance most file systems
recommend 10% free at all times although I doubt that much is really
required.
But with the above setting you should be able to fly fairly close to the
wire as it
will always use the most empty disk.
Updates listed on http://netwinsite.com/dnews/updates.htm
ChrisP.
