Re: [Tracker] trackerd consumes over 6 MB of RAM



David Nielsen wrote:
ons, 15 11 2006 kl. 07:12 +0200, skrev Tshepang Lekhonkhobe:
Tracker, currently indexing, consumes 18 MB of RAM, so you should
probably remove the 3-6 MB consumption advertisement. It currently
indexes about 200GB  of data of about 52,000 files.

here's the output of du .Tracker/
1880    .Tracker/databases/Files/0001
1892    .Tracker/databases/Files/0002
1916    .Tracker/databases/Files/0003
1896    .Tracker/databases/Files/0004
7592    .Tracker/databases/Files
56712   .Tracker/databases
69992   .Tracker/

Bah that's nothing I ran a test of Tracker yesterday on my FC6 setup
since Tracker got put into Fedora Extras a few days ago. I hit 20megs
easily. It looks to me as it might be leaking a bit since it consumes
more and more over time.

you should try using --enable-low-memory with trackerd which lowers all cache values but gives you slower indexing. If this still produces high leves of memory usgae then we can be sure its memory leaking (it will grow slightly in memory due to increased no of watches + we are zipping full text which can take double the memory of the contents you are zipping)

we by default use a higher level of caching with sqlite using on average up to 3MB per connection per thread (max of 4 threads with 3 connections each means theroectically max memory usage by sqlite would be 36 MB)

I am not sure what benefits this cache gives over the systems disk cache but I will have a play. It will need some for sorting and grouping results but it could do with some experimentation


I'll build some tracker enabled nautilus rpms we we can get that tested
on Fedora as well.

mclasen told me he is building an external lib for nautilus that encapsulates tracker and beagle interfaces for searching in nautilus. It will be in FC7 and he hopes to get it upstream into nautilus proper (as he is a redhat guy and the main gtk hacker that seems likely)


--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/




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