allow to disable automatic stat

Issue #11 wontfix
Christoph Anton Mitterer
created an issue

Hi.

It seems that it's possible to cause as statistics update by sending USR2 to the db process.

For that reason, it would IMHO be a good idea if responsibility for that is generally put into cron. At least it would be nice if one could disable stat from within sks, e.g. -stat none or something like that, so that I can do A crontab like 0 * * debian-sks pkill -F /run/sks/sksdb.pid -USR2 sks || exit 1 without doing a double stat at 03:00

Cheers, Chris.

See also: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=718608

Comments (4)

  1. John Clizbe

    Thank you for explaining it to me. I added it to SKS.

    There is no requirement that USR2 be used to periodically generate stats, just as there is no requirement that it be done at the top of the hour, overlapping the setting for statshour. Indeed, when it was originally added, it was set to trigger at 40 minutes past the hour on the hours that Peter Pramberger and Kristen's sites collected stats.

    IMNSHO, allowing for disabling stats in their entirety except by signally USR2 is Not A Good Idea. We still have folks who do not enable initial_stat.

    Fix the cron entry.

  2. Christoph Anton Mitterer reporter

    Thank you for explaining it to me. I added it to SKS. Hmm? What do you mean exactly?

    Well the only reason for that idea was basically what Kristian mentioned on the sks-devel list just yesterday: Many servers do it via cron nowdays... some not. Which leads to the strange effect that servers are weighted differently though they actually have about the same number of keys.

    When I now do a cron entry for say hourly, many people will use, e.g.: 0 * * * * pkill -USR2 sks || exit 1 But that leads to a stat being unnecessarily done twice at (per default) 03:00.

    My idea would of course have included, that packages should a cron file per default, so that all servers are still updated :-)

    Anyway... was just an idea,... so if you think the stat trigger from within the program is still neccesary, just close that ticket!

  3. Log in to comment