Git-Mirrors/brozzler
1
0
Fork 0
mirror of https://github.com/internetarchive/brozzler.git synced 2025-02-24 08:39:59 -05:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix bug claiming site, looks like there could be a race condition with other worker claiming the same site

Browse Source
This commit is contained in:
Noah Levitt 2015-09-04 01:36:29 +00:00
parent 3c23aa8fd4
commit 5fe2805285
4 changed files with 10 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
brozzler/__init__.py
View File

@ -41,6 +41,7 @@ class ReachedLimit(Exception):
class Rethinker:
import logging
logger = logging.getLogger(__module__ + "." + __qualname__)
def __init__(self, servers=["localhost"], db=None):
@ -50,8 +51,8 @@ class Rethinker:
# https://github.com/rethinkdb/rethinkdb-example-webpy-blog/blob/master/model.py
# "Best practices: Managing connections: a connection per request"
def _random_server_connection(self):
import rethinkdb as r
import random
import rethinkdb as r
while True:
server = random.choice(self.servers)
try:
@ -66,6 +67,7 @@ class Rethinker:
time.sleep(0.5)
def run(self, query):
import rethinkdb as r
while True:
with self._random_server_connection() as conn:
try:

4
brozzler/frontier.py
View File

@ -85,7 +85,7 @@ class RethinkDbFrontier:
result = self.r.run(r.table("sites")
.between(["ACTIVE",False,0], ["ACTIVE",False,250000000000], index="sites_last_disclaimed")
.order_by(index="sites_last_disclaimed").limit(1).update({"claimed":True},return_changes=True))
self._vet_result(result, replaced=[0,1])
self._vet_result(result, replaced=[0,1], unchanged=[0,1])
if result["replaced"] == 1:
site = brozzler.Site(**result["changes"][0]["new_val"])
else:
@ -157,7 +157,7 @@ class RethinkDbFrontier:
for url in outlinks:
if site.is_in_scope(url, parent_page):
if brozzler.is_permitted_by_robots(site, url):
new_child_page = brozzler.Page(url, site_id=site.id, hops_from_seed=parent_page.hops_from_seed+1)
new_child_page = brozzler.Page(url, site_id=site.id, hops_from_seed=parent_page.hops_from_seed+1, via_page_id=parent_page.id)
existing_child_page = self.get_page(new_child_page)
if existing_child_page:
existing_child_page.priority += new_child_page.priority

3
brozzler/site.py
View File

@ -81,13 +81,14 @@ class Site(brozzler.BaseDictable):
return False
class Page(brozzler.BaseDictable):
def __init__(self, url, id=None, site_id=None, hops_from_seed=0, redirect_url=None, priority=None, claimed=False, brozzle_count=0):
def __init__(self, url, id=None, site_id=None, hops_from_seed=0, redirect_url=None, priority=None, claimed=False, brozzle_count=0, via_page_id=None):
self.site_id = site_id
self.url = url
self.hops_from_seed = hops_from_seed
self.redirect_url = redirect_url
self.claimed = bool(claimed)
self.brozzle_count = brozzle_count
self.via_page_id = via_page_id
self._canon_hurl = None
if priority is not None:

6
cluster-control.sh
View File

@ -91,7 +91,7 @@ _reset() {
tstamp=$(date +"%Y%m%d%H%M%S")
echo "renaming rethinkdb database archiveit_brozzler to archiveit_brozzler_$tstamp"
python3.4 <<EOF
PYTHONPATH=/home/nlevitt/tmp/brozzler-venv/lib/python3.4/site-packages python3.4 <<EOF
import rethinkdb as r
with r.connect("wbgrp-svc035") as conn:
r.db("archiveit_brozzler").config().update({"name":"archiveit_brozzler_$tstamp"}).run(conn)
@ -111,11 +111,11 @@ _start() {
set -e
set -x
/home/nlevitt/tmp/brozzler-venv/bin/warcprox --dir=/1/brzl/warcs --rethinkdb-servers=wbgrp-svc020,wbgrp-svc035,wbgrp-svc036 --rethinkdb-db=archiveit_brozzler --rethinkdb-big-table --cacert=/1/brzl/warcprox-ca.pem --certs-dir=/1/brzl/certs --address=0.0.0.0 --base32 --gzip --rollover-idle-time=180 &>/1/brzl/logs/warcprox.out &
PYTHONPATH=/home/nlevitt/tmp/brozzler-venv/lib/python3.4/site-packages:/home/nlevitt/workspace/brozzler:/home/nlevitt/workspace/warcprox:/home/nlevitt/workspace/ait5 /home/nlevitt/tmp/brozzler-venv/bin/warcprox --dir=/1/brzl/warcs --rethinkdb-servers=wbgrp-svc020,wbgrp-svc035,wbgrp-svc036 --rethinkdb-db=archiveit_brozzler --rethinkdb-big-table --cacert=/1/brzl/warcprox-ca.pem --certs-dir=/1/brzl/certs --address=0.0.0.0 --base32 --gzip --rollover-idle-time=180 --kafka-broker-list=qa-archive-it.org:6092 --kafka-capture-feed-topic=ait-brozzler-captures &>/1/brzl/logs/warcprox.out &
sleep 5
/home/nlevitt/workspace/ait5/scripts/brozzler-job-starter.py &> /1/brzl/logs/ait-job-starter.out &
PYTHONPATH=/home/nlevitt/tmp/brozzler-venv/lib/python3.4/site-packages:/home/nlevitt/workspace/brozzler:/home/nlevitt/workspace/warcprox:/home/nlevitt/workspace/ait5 /home/nlevitt/workspace/ait5/scripts/brozzler-job-starter.py &> /1/brzl/logs/ait-job-starter.out &
sleep 5