Haystack index update script + cronjob

written on Sunday, May 8, 2011

Yukmarks doesn't really have any users except for me and my girlfriend, so updating the search index manually has never been very difficult. Just punch up my fabfile and run update_search(). Trivial.

However, I am seriously, profoundly lazy, so all those keystrokes were getting annoying. Plus, manually updating search is just stupid & inefficient. I could use the same amount -- or less -- of keystrokes to just Ctrl-F on my Yukmarks profile page. So I wrote a script to automatically update my Haystack search. A cron job runs it every 15 minutes.

The code

import subprocess
import sys
import logging

domain_dir = "/a/mattdeboard.net/"
appdir = domain_dir + "src/yukproj/"
whoosh_dir = appdir + "yuk/whoosh/"

def update():
                        format='%(asctime)s %(levelname)s:%(message)s',
                        datefmt='%m/%d/%Y %H:%M:%S')
    logging.info('Starting index update.')
    update_index = subprocess.Popen(['sudo', '-u', 'www-data',
                                   appdir+'manage.py', 'update_index'],
    apachereload = subprocess.Popen(['sudo',
    if not any((update_index.returncode, apachereload.returncode)):
        logging.info('Index successfully updated.')
        subs = [update_index, apachereload]
        logging.error('**INDEX UPDATE FAILED**')
        logging.error('The following exit codes were returned:')
        logging.error('- update_index: %s' % update_index.returncode)
        logging.error('- apachereload: %s' % apachereload.returncode)
        for sub in subs:
            if sub.returncode:
                logging.error('Error information:')
                logging.error('stdout: %s' % sub.communicate()[0])
                logging.error('stderr: %s' % sub.communicate()[1])

if __name__ == '__main__':

The (root) cron job:

0,15,30,45 * * * * /a/mattdeboard.net/bin/python /a/mattdeboard.net/src/yukproj/ \\
srchupdate.py -c|mail -s "Search Update Complete" matt

(I have it all on one line in crontab, but broken up into two here for ease of reading.)

What do

So basically every 15 minutes, the server runs srchupdate.py, and logs the results to a log file outside the project directory. If the update fails, it logs the exit status values, stderr and stdout data (using the communicate() method and returnvalue attribute of Python's excellent subprocess.Popen). This captures traceback info and has made debugging much easier.

The biggest source of friction is Whoosh, which powers my Haystack install. Whoosh is pure Python, and very easy to install. However, it is extremely slow. I'd probably even say ponderous. For ~350 bookmark entries on Yuk, it takes about 10 seconds to update. From what I understand, Solr is much faster, but has a much steeper learning curve. For Yukmarks, I think Whoosh is fine, but I doubt going forward I'd use it for any serious projects where speed is important.

If you got here after Googling for a Haystack auto-update solution, I hope this helps.

Edit: Removed a paragraph about incremental indexing. Haystack does this by default.

This entry was tagged django, haystack, python and whoosh