[recoll-user] peaks in IO that freeze the machine

  • From: Ramon Diaz-Uriarte <rdiaz02@xxxxxxxxx>
  • To: recoll-user@xxxxxxxxxxxxx
  • Date: Fri, 02 Sep 2011 12:26:13 +0200

Dear All,

I've been a very happy user of recoll. Lately, however, I've started to
notice that I periodically get peaks in IO that freeze my systems.


Details:

- I index my home directory (my recoll.conf is at the bottom) using real
  time indexing. (I also monitor /usr/share/doc with another recollindex,
  but that rarely causes trouble).       

- My current recoll version is 1.15.9-1 (as packed by Debian). 

- I am monitoring IO usage with iotop. 

- From time to time (e.g., when I delete/refile messages in my Maildir, or
  when offlineimap deletes messages) I can see recollindex doing lots of
  IO (iotop shows up to 99%). 

- This seems to lead to many other processes being left waiting (as shown
  by top).  For instance, while answering email, Emacs freezes and you
  cannot see what your are typing. Or the command line freezes if you are
  doing a simple "ls".

- This can last from barely less than a second to a few seconds, but it is
  very intrusive.

- The log output shows all sorts of messages, many of them "No such file
  or directory", but not all. In particular, it does not seem related (as
  far as I can tell) to processing some very huge text file.

- This problem happens in several machines, from a tiny Asus netbook to a
  quad-core workstation. All run Debian GNU/Linux and recent stock Debian
  kernels (e.g., 2.6.39).




I am attaching a screen capture that shows (left to right, top to bottom):
  * tail -f recolltrace
  * iotop -o
  * iotop -o -a
  * top

it shows a case where IO from recollindex seems to do lots of IO and a
large fraction of processes (23%, but in other cases as many as 50%) are
waiting, I think because of IO.


What am I doing wrong? Is there a way to avoid these problems?


Thanks a lot, and thanks for such a great program.

Best,

R.

P.S. My ~/.recoll/recoll.conf

****************
indexstemminglanguages = english spanish
skippedNames = *.RData *.Rdata *~ .beagle .bzr .git .hg .svn .thumbnails CV\
S Cache bin cache* caughtspam loop.ps recollrc tmp  .xsessio\
n-errors .recoll* xapiandb recollrc #*
skippedPaths = /home/ramon/.local/share/data /home/ramon/.wuala /home/ramon\
/WualaDrive /home/ramon/caca /home/ramon/tmp
topdirs = /home/ramon
## /usr/share/doc  indexed in a separate process


loglevel = 3
logfilename = /home/ramon/.recoll/recolltrace

textfilemaxmbs = 100

****************

PNG image

 



-- 
Ramon Diaz-Uriarte
Department of Biochemistry
Universidad Autónoma de Madrid
Spain

http://ligarto.org/rdiaz

Temporarily at:
Structural Biology and Biocomputing Programme
Spanish National Cancer Centre (CNIO)
 
Phone: +34-91-732-8000 ext. 3019
Fax: +-34-91-224-6972

Other related posts: