-----Original message----- > From:Joe Zhang <[email protected]> > Sent: Thu 29-Nov-2012 23:15 > To: [email protected] > Subject: Re: Nutch efficiency and multiple single URL crawls > > I realize I might be asking a real naive question here :)
no problem > > Why can't we put all the base URLs in a single seed file, and use a same > config file (which has all the filtering patterns) to do the crawl? Is > there anything wrong with this approach? No, there's nothing wrong in doing that. You can use the FreeGenerator tool to generate a fetch list from a single (or multiple) seed file(s). But what are you trying to accomplish? Something large scale (tens of millions) or much larger or smaller? Usually concerns of what is going to be crawled only matters if you're doing something on a large or massive scale. > > On Thu, Nov 29, 2012 at 2:46 PM, Alejandro Caceres < > [email protected]> wrote: > > > Got it, I will try that out, that's an excellent feature. Thank you for the > > help. > > > > > > On Thu, Nov 29, 2012 at 4:06 AM, Markus Jelsma > > <[email protected]>wrote: > > > > > As i said, you don't rebuild, you just overwrite the config file in the > > > hadoop config directory on the data nodes. Config files are looked up > > there > > > as well. Just copy the file to the data nodes. > > > > > > -----Original message----- > > > > From:AC Nutch <[email protected]> > > > > Sent: Thu 29-Nov-2012 05:38 > > > > To: [email protected] > > > > Subject: Re: Nutch efficiency and multiple single URL crawls > > > > > > > > Thanks for the help. Perhaps I am misunderstanding, what would be the > > > > proper way to leverage this? I am a bit new to Nutch 1.5.1, I've been > > > using > > > > 1.4 and have generally been using runtime/deploy/bin/nutch with a .job > > > > file. I notice things are done a bit differently in 1.5.1 with the lack > > > of > > > > a nutch runtime and nutch deploy directories. How can I run a crawl > > while > > > > leveraging this functionality and not having to rebuild the job file > > each > > > > new crawl? More specifically, I'm picturing the following workflow... > > > > > > > > (1) update config file to restrict domain crawls -> (2) run command > > that > > > > crawls a domain with changes from config file while not having to > > rebuild > > > > job file -> (3) index to Solr > > > > > > > > What would the (general) command be for step (2) is my question. > > > > > > > > On Mon, Nov 26, 2012 at 5:16 AM, Markus Jelsma > > > > <[email protected]>wrote: > > > > > > > > > Hi, > > > > > > > > > > Rebuilding the job file for each domain is not a good idea indeed, > > > plus it > > > > > adds the Hadoop overhead. But you don't have to, we write dynamic > > > config > > > > > files to each node's Hadoop configuration directory and it is picked > > up > > > > > instead of the embedded configuration file. > > > > > > > > > > Cheers, > > > > > > > > > > -----Original message----- > > > > > > From:AC Nutch <[email protected]> > > > > > > Sent: Mon 26-Nov-2012 06:50 > > > > > > To: [email protected] > > > > > > Subject: Nutch efficiency and multiple single URL crawls > > > > > > > > > > > > Hello, > > > > > > > > > > > > I am using Nutch 1.5.1 and I am looking to do something specific > > with > > > > > it. I > > > > > > have a few million base domains in a Solr index, so for example: > > > > > > http://www.nutch.org, http://www.apache.org, > > > http://www.whatever.cometc. I > > > > > > am trying to crawl each of these base domains in deploy mode and > > > retrieve > > > > > > all of their sub-urls associated with that domain in the most > > > efficient > > > > > way > > > > > > possible. To give you an example of the workflow I am trying to > > > achieve: > > > > > > (1) Grab a base domain, let's say http://www.nutch.org (2) Crawl > > the > > > > > base > > > > > > domain for all URLs in that domain, let's say > > > http://www.nutch.org/page1 > > > > > , > > > > > > http://www.nutch.org/page2, http://www.nutch.org/page3, etc. etc. > > > (3) > > > > > store > > > > > > these results somewhere (perhaps another Solr instance) and (4) > > move > > > on > > > > > to > > > > > > the next base domain in my Solr index and repeat the process. > > > Essentially > > > > > > just trying to grab all links associated with a page and then move > > > on to > > > > > > the next page. > > > > > > > > > > > > The part I am having trouble with is ensuring that this workflow is > > > > > > efficient. The only way I can think to do this would be: (1) Grab a > > > base > > > > > > domain from Solr from my shell script (simple enough) (2) Add an > > > entry to > > > > > > regex-urlfilter with the domain I am looking to restrict the crawl > > > to, in > > > > > > the example above that would be an entry that says to only keep > > > sub-pages > > > > > > of http://www.nutch.org/ (3) Recreate the Nutch job file (~25 > > sec.) > > > (4) > > > > > > Start the crawl for pages associated with a domain and do the > > > indexing > > > > > > > > > > > > My issue is with step #3, AFAIK if I want to restrict a crawl to a > > > > > specific > > > > > > domain I have to change regex-urlfilter and reload the job file. > > > This is > > > > > a > > > > > > pretty significant problem, since adding 25 seconds every single > > > time I > > > > > > start a new base domain is going to add way too many seconds to my > > > > > workflow > > > > > > (25 sec x a few million = way too much time). Finally the > > > question...is > > > > > > there a way to add url filters on the fly when I start a crawl > > and/or > > > > > > restrict a crawl to a particular domain on the fly. OR can you > > think > > > of a > > > > > > decent solution to the problem/am I missing something? > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > ___ > > > > Alejandro Caceres > > Hyperion Gray, LLC > > Owner/CTO > > >
