Fedora14 is filling up my HDD without a reason

[Joel Rees]

You know, I'm sitting here trying to figure out what exactly you're
trying to say, and I thought about the diference between clean
metadata and clean all.

On Thu, May 12, 2011 at 9:17 PM, Tim <ignored_mailbox at yahoo.com.au> wrote:
> On Thu, 2011-05-12 at 20:46 +0900, Joel Rees wrote:
>> Yesterday's metadata: 15M
>> Yesterday's downloads: 90M
>>
>> 17% additional bandwidth burden.
>>
>> Today, no downloads, so it would be nothing but overhead.
>>
>> I'm not sure, since the mirrors I usually connect to are on gigabit
>> pipes (and my pipe is limited at 1Mbit), that 15 Megabytes (in about a
>> minute and a half) three or four times a week constiutes an
>> unreasonable burden on the infrastructure.

So I decided to add a little to today's burden on my local mirrors and
check the difference on a day with no downloads.

No difference. So, I was wrong. No extra overhead for clean all vs.
clean metadata on days with no downloads.

> Multiplied by all the other people who would do the same...

0 multiplied by what, a thousand? A million? (I wish. We wish.)

> Let's not just consider /that/ server, but whether every server was
> overburdened by lots of users needlessly adding to their traffic.

What is the extra burden?

If I have, say, 90M of updated packages, are you saying that having
the old packages in my cache somehow saves bandwidth? Has yum been
upgraded to run in diff mode, then? That would be good news, indeed,
although I haven't seen such evidence in the download sizes.

Now, if I had more than one machine in the house running the same
version of Fedora, I would definitely be looking at putting up alocal
proxying miror of some sort. I see from other comments that some
people seem to be making use of their cache as manual mirrors, ...

>> I have had my cache clogged on occasion, preventing security updates
>> from downloading, which is why I tend to use the clean all option.
>
> Though, rarely does the fault actually require the brute force "all"
> option.

Well, anyway, since you insist there's an advantage, I'll dig around
and see what it might be. Right now, I'm afraid you're either not
making sense, or you're assuming I have more than one machine running
the same distribution on the inside LAN. Assuming I'm running in a
hosted VM, maybe?

Joel Rees