Re: /var 100% used error

[Jeff Johnson <n3npq jbj gmail com>]

On Apr 11, 2007, at 11:23 AM, Atul Ankola wrote:

> Jeff,
>
> Filling up /var wasn't done on purpose - it happened on a telecom  
> server running in the field due to another problem.

Understood.

I suppose its time for rpm to guesstimate its disk space needs on / 
var and try
to fail gracefully if it looks like there is insufficient space on /var.

The easiest approach would be to automatically check manually configured
minimum free space threshholds for mounted file systems before  
attempting
anything. All that is hard there is committing to a configuration  
format and writing
a bleeping parser.

A little trickier but quite doable is to estimate the space needed  
for rpmdb entries.
The tricky part is that there's no way to do the arithmetic  
precisely, if -- for no other
reason -- that information returned from statfs is insufficient  
(because of root reserved
space on file systems) to do the arithmetic. The current disk  
accounting multiplies
by 21/20 (i.e. assumes 5% root reserved space) which is (or was) the  
ext[23] default.

> The --ignore size was used because our preinstall script was  
> clearing the installation partition which doesn't have enough room  
> until our preinstall script finishes.
>
> Ok, we're still on 4.4.2.  I'll see if we can upgrade to 4.4.8 or  
> later.

Glad to have you along. Try <rpm-devel lists dulug duke edu> if you  
need help.

Backporting the stale lock detection/correction from rpm-4.4.8-0.4 to  
PNAEL's rpm-4.4.2 fork
is quite doable if you want/need that instead.

73 de Jeff