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

[Popcon-developers] Bug#298760: popularity-contest: Gives inaccurate information when disk is mounted with noatime option

On Thu, Aug 04, 2005 at 04:58:53PM -0500, Nathan Stratton Treadway wrote:
> A related issue is the case where a program such as "integrit" updates
> the atime of all files in /bin and /usr (and elsewhere) each day.  

> So, my point is that if you add special handling for the noatime case,
> you might want to also handle the "atimes are all too close together"
> problem at the same time.

For what it's worth, I discovered that a similar problem affected
another machine, one that does not run "integrit".  In this case, it was
backing up the system (e.g. to tape) that set all of my atimes to be no
older that a few days, and thus all packages are still getted "voted".

It's clear that there are lots of situations that make the votes
inaccurate, so you (popcon-developers) may just decide to leave things
as they are.  However, it occurs to me that one possible approach is to
define a new field in the report header, something like "ATIME_STATUS:"
or whatever.  When set to some sort "invalid" status, this would
indicate to the server that this particular report should be included in
the installed count but not the vote count, or whatever.

In my two situations, the client script would decide to set this flag
based on the fact that all the atimes are within a short time span of
each other but the ctimes are more spread out.  (Obviously the trick to
to pick reasonable cutoff values for this calculation...)

In the case described by the original submitter, this status could get
set based on the existence of a noatime-mounted partition, for example.
(Or perhaps even by noticing that dpkg-query, perl-base, etc. had atimes
that were clearly too old given that popularity-contest itself uses

Anyway, something to consider.


Nathan Stratton Treadway  -  nathanst@ontko.com  -  Mid-Atlantic region
Ray Ontko & Co.  -  Software consulting services  -   http://www.ontko.com/
 GPG Key: http://www.ontko.com/~nathanst/gpg_key.txt   ID: 1023D/ECFB6239
 Key fingerprint = 6AD8 485E 20B9 5C71 231C  0C32 15F3 ADCD ECFB 6239
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/popcon-developers/attachments/20050807/0ff852b3/attachment.pgp

Reply to: