Re: Quota problem
On Tue, 3 Oct 2000, Michael Meskes wrote:
> The way the quota package installs itself right now is not the best option,
> since it checks quota in the foreground on a fresh installation. Now this
> could be solved by moving quotacheck into the background. This surely would
mmh, is that really so good? maybe your package is not the only one
beeing installed and system partitions might have also quotas (obviously
nonesense, but who knows), so quotas might get screwed up when quotas was
configured like this prior installing your package.
> solve the installation problem, but how about the check during boot time?
I would say there it should stay even on huge machines. I think that if
you have the money to buy one huge box, it would be better to buy 2 and
make use high-availability software.
> Some systems will need an hour or so to just check quota, but moving this
> into the background as well may lead to incorrect data, since changes to the
> filesystem during a quotacheck may prove incorrect results.
I think the only way to cope with that would result in rewriting the
> One idea was to only check quota if fsck returns an error, but that would
> mean that the we only get correct numbers after an incorrect umount. Besides
> this would also mean that the call to quotacheck would have to be moved to
> sysvinit's scripts.
As you say this is definitly not the right way.