The following reply was made to PR general/1376; it has been noted by GNATS.
From: TTSG <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED] (Dean Gaudet)
Cc: [EMAIL PROTECTED], [EMAIL PROTECTED]
Subject: Re: general/1376: Server stops servicing requests when log disk is full
Date: Sun, 9 Nov 1997 09:53:51 -0500 (EST)
>
> On Sat, 8 Nov 1997, TTSG wrote:
>
> > TransferLog "|/usr/local/bin/cronolog
> > /usr/local/etc/httpd/logs/%Y%m%d/access.log"
>
> Ah, that pretty much explains it. There is a 4k pipe buffer, when it's
> full all tasks writing the pipe are blocked. My guess is that cronolog is
> not reading its input when the disk is full. You'll have to get cronolog
> fixed -- it has to continue reading its input even when the disk is full.
>
Ok Dean, thanks. I didn't see this before I wrote the other one.
This explains it completely. Thanks.
Andrew, is there something you can do if the disk is full to just
/dev/null the output until its time to rotate the logs. Then if there is
room free at that time continue, else /dev/null it again?
Thanks Dean and Andrew.
Tuc/TTSG
