update-apt-xapian-index is run both daily and weekly

[Tormod Volden]

Binary package hint: apt-xapian-index

/etc/cron.weekly/apt-xapian-index runs it weekly. In addition it is run daily inside /etc/cron.daily/apt

With the fixes in 0.37 in Maverick, it has become less of a catastrophe but still it is one of the most annoying things on the Ubuntu desktop IMHO: 2 minutes of full CPU, taking 100-200MB, kicking in when you least need it.

[Tim Schutte (kc8hr)]

apt-xapian-index crashed a program running under Wine, FullTiltPoker.exe--and I was winning! ;-(

[Tormod Volden (tormodvolden)]

Additionally, whereas the weekly script checks for virtual machines (for ionice) and uses the --update option and nice, the daily script does not.

[aexl (aexl)]

just took a look into the maverick package: http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/maverick/apt/maverick/annotate/head:/debian/apt.cron.daily

things seem to be better there:
nice ionice -c3 update-apt-xapian-index -q

but we have the same issue as in bug 363695 - it should be something like this for best niceness:
nice -n 19 ionice -c 3 update-apt-xapian-index --update --quiet

authors file: http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/maverick/apt/maverick/annotate/head:/AUTHORS

[StefanHuszics (stefan-huszics)]

Is there a bug open somewhere that puts in question the entire existance of this update script as a cronjob? Please someone direct me to it in that case.

On my laptop, where automatic updates are completely turned off it's just insane that this cronjob updates are run at all, ever. It should be run in conjunction with updating or installing a package (ie when the package list has atually changed), not at preset intervals.

[Tormod Volden (tormodvolden)]

Stefan, AFAIK no, so please file one.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Sérgio Faria (sergio91pt)]

On apt 0.8.16~exp5ubuntu13, theres this:
[ Adam Conrad ]
* On armel, call update-apt-xapian-index with '-u' to keep the CPU
  and I/O usage low. We would do this on all arches, but there's a
  regression risk here, but that's better than killing slow systems.

Shouldn't the daily script run with -u *always*, and let the re-built exclusively for the weekly script?