Regarding the machine, you could run it on anything, it all depends
what kind of performance you want. So yes you could run it on the
machine that you have or something like the linux machine that I
have. And the DSL connection should be fine, you just need to make
sure that it stays up the whole time, because if not it will start
erring out, and you will have to re-fetch that whole segment as there
is no way to pick up from where you left off.
The only reason that I merged the segments was that I had many of
them, and I wanted to build a big one before I started creating new
ones, another advantage of merging is that you can use that to clear
out unwanted urls. For example I had a bunch of .js files in there,
that I didn't want to have as part of the index, so I cleared them out.
I used "bin/nutch mergesegs".
Regarding merging the other parts, I have never used, but I don't
think that it is necessary unless you have multiple linkdbs, etc. in
my case I do not.
-John
On Mar 3, 2009, at 7:14 PM, [email protected] wrote:
Hi,
I will need to index all links in domains then. What do you think a
linux box like yours with DSL connection is OK to index the domains
I have?
Why only segments? I thought we need to merge all sub folders under
crawl folder. What did you use for merging them?
Thanks.
A.
-----Original Message-----
From: John Martyniak <[email protected]>
To: [email protected]
Sent: Tue, 3 Mar 2009 3:21 pm
Subject: Re: what is needed to index for about 10000 domains
Well the way that nutch works is that you would inject your list of
domains into the DB, and that would be the starting point. Since
nutch uses a crawler it would grab those pages, and determine if
there are any links on those pages, and then add them to the DB. So
the next time that you generated your urls to fetch, it would take
your original list, plus the ones that it found to generate the new
segment.?
?
If you wanted to limit it to only pages contained on your 10000
domains, you could use the regex-urlfilter.txt file in the conf
directory to limit it to your list. But you would have to create a
regular expression for each one.?
?
I am not familiar with the merge script on the wiki, but have merged
segments before and it did work. But that was on Linux, don't think
that should make a difference though.?
?
-John?
?
?
On Mar 3, 2009, at 5:10 PM, [email protected] wrote:?
?
?
Hi,?
?
Thanks for the reply. I have list? of those domains only. I am not
> sure how many pages they have. Is a DSL? connection sufficient to
> run nutch in my case. Did you run nutch for all of your pages at
> once or separately for a given subset of them. Btw, yesterday I >
tried to use merge shell script that we have on wiki. It gave a lot
> of errors. I run it on cygwin though.?
?
Thanks.?
A.?
?
?
?
?
?
?
?
-----Original Message-----?
From: John Martyniak <[email protected]>?
To: [email protected]?
Sent: Tue, 3 Mar 2009 1:44 pm?
Subject: Re: what is needed to index for about 10000 domains?
?
?
?
?
?
?
?
?
?
I think that in order to answer that questions, it is necessary to
> know how many total pages are being indexed.??
??
?
I currently have ~3.5 million pages indexed, and the segment >
directories are around 45GB, The response time is relatively fast.??
??
?
In the test site it is running on a dual processor Dell 1850 with >
3GB of RAM.??
??
?
-John??
??
?
On Mar 3, 2009, at 3:44 PM, [email protected] wrote:??
??
?
Hello,??
?
??
?
I use nutch-0.9 and need to index about 10000? domains.? I want to
>> > know? minimum requirements to hardware and memory.??
?
??
?
Thanks in advance.??
?
Alex.??
??
?
?
?
?
?
?
