[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

[Popcon-developers] Bug#462268: Bug#462268: popcon: stay out of lenny for a while



On Fri, Mar 14, 2008 at 04:52:17PM +0100, Petter Reinholdtsen wrote:
> The popcon version graph on <URL:http://popcon.debian.org/> has
> flattened out, and the increase for stable (1.41) since version 1.43
> was uploaded to unstable is almost the same as the amount of hosts
> registered using unstable (~5000).  This make me suspect that the DOS
> issue might be the reason we do not see more submissions registered.
> On the other hand, the submissions in stable did not increase as the
> exact same date as the upload was done to unstable.  This make me more
> unsure if the increase is related to the unstable upload or not.

I asked for feedback from Holger about the DOS issue and got nothing. So
I rather thing the DOS issue was temporary due to external network issue
and not a real issue we need to adress. At least I will assume that
until I get real data. The stable submissions increased even while we
made no change to unstable.

> Perhaps it is time to allow version 1.43 into testing, and see if the
> amount of stable submissions increases with that amount (~14700)?

I really do not like the hack in 1.43 and I do not think it belong in
testing. The number of unstable submission is not encouraging.

> Anyone working on a different approach to avoid all clients in a given
> time zone contacting the popcon server at the same time?

I have the following proposal:

We move the popcon cronjob to cron.dayly, and pick a random weekday for
each submitter. The dayly cronjob check if it is running on this weekday
before reporting, else it abort immediatly.

This spread the load on the whole week  (this reduce network bandwidth
issue more that the 1.43 scheme) and does not require a cron.d entry.

Cheers,
-- 
Bill. <ballombe at debian.org>

Imagine a large red swirl here. 





Reply to: