December 2009
November 2009
October 2009
September 2009
June 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
July 2008
June 2008
October 2007
September 2007
Our Solarwinds Network Performance Monitor has a problem rendering custom reports on occasion. For something like that, there isn't an existing plugin for Nagios. Writing these plugins is easy. All there is to it is exit statuses. After reading this, you should have an idea of how to write a Nagios plugin for a variety of web applications.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 | #!/usr/bin/env python from mechanize import Browser from optparse import OptionParser # Exit statuses recognized by Nagios UNKNOWN = -1 OK = 0 WARNING = 1 CRITICAL = 2 def open_url(br, url): """Use a given mechanize.Browser to open url. If an exception is raised, then exit with CRITICAL status for Nagios. """ try: response = br.open(url) except Exception, e: # Catching all exceptions is usually a bad idea. We want to catch # them all to report to Nagios here. print 'CRITICAL - Could not reach page at %s: %s' % (url, e) raise SystemExit, CRITICAL return response # I'm going to be using optparse.OptionParser from now on. It makes # command-line args a breeze. parser = OptionParser() parser.add_option('-H', '--hostname', dest='hostname') parser.add_option('-u', '--username', dest='username') parser.add_option('-p', '--password', dest='password') parser.add_option('-r', '--report_url', dest='url', help="""Path to report relative to root, like /NetPerfMon/Report.asp?Report=Hostname+__+IPs""") parser.add_option('-v', '--verbose', dest='verbose', action='store_true', default=False) parser.add_option('-q', '--quiet', dest='verbose', action='store_false') options, args = parser.parse_args() # Check for required options for option in ('hostname', 'username', 'password', 'url'): if not getattr(options, option): print 'CRITICAL - %s not specified' % option.capitalize() raise SystemExit, CRITICAL # Go to the report and get a login page br = Browser() report_url = 'https://%s%s' % (options.hostname, options.url) open_url(br, report_url) br.select_form('aspnetForm') # Solarwinds has interesting field names # Maybe something with asp.net br['ctl00$ContentPlaceHolder1$Username'] = options.username br['ctl00$ContentPlaceHolder1$Password'] = options.password # Attempt to login. If we can't, tell Nagios. try: report = br.submit() except Exception, e: print 'CRITICAL - Error logging in: e' % e raise SystemExit, CRITICAL report_html = report.read() # class=Property occurs in every cell in a Solarwinds report. If it's not # there, something is wrong. if 'class=Property' not in report_html: print 'CRITICAL - Report at %s is down' % report_url raise SystemExit, CRITICAL # If we got this far, let's tell Nagios the report is okay. print 'OK - Report at %s is up' % report_url raise SystemExit, OK |
To use our plugin, we need to do a bit of Nagios configuration. First, we need to define a command.
define command{
command_name check_npm_reports
command_line /usr/local/bin/reportmonitor.py -H $HOSTADDRESS$ $ARG1$
}
After that, we define a service.
define service{
use generic-service
host_name solarwinds-server
service_description Solarwinds reports
check_command check_npm_reports!-u nagios -p some_password -r '/NetPerfMon/Report.asp?Report=Hostname+__+IPs'
}
Doing anything with SOAP is a pain without a WSDL, which is the case with Zimbra. All of the Howtos I found about SOAP and ruby either required a WSDL or making several classes in a special, undocumented way to trick a SOAP::RPC::Driver instance into working. Both were unacceptable. After much hardship, I found an easier to read way to do SOAP without an WSDL in ruby, by building SOAP::Elements myself. Here is the code, documented to be easy to read, use, and extend.
# Incomplete library for interacting with Zimbra # # require 'zimbra' # # host = 'zimbra.tylerlesmann.com' # user = 'root' # passwd = 'hard_password' # creds = Zimbra.authenticate(host, user, passwd) # usercreds = Zimbra.masquerade(host, creds.authToken, 'tlesmann') # Zimbra.createappointment(host, usercreds.authToken, # Time.local(2009, 6, 26), 'Make a blog post', 'Maybe some Java', [ # '/home/tlesmann/Documents/java.png', # '/home/tlesmann/Documents/tutorial.pdf', # ]) require 'net/http' require 'net/https' require 'soap/element' require 'soap/rpc/driver' require 'soap/processor' require 'soap/streamHandler' require 'soap/property' require 'zimbra/multipart' module Zimbra # Builds and sends AuthRequest to a provided Zimbra host. # # Returns a SOAP::Mapping instance, with an authToken attribute def self.authenticate(host, name, password) header = SOAP::SOAPHeader.new body = SOAP::SOAPBody.new(element('AuthRequest', nil, { 'xmlns' => 'urn:zimbraAdmin', }, [ element('name', name), element('password', password), ] )) envelope = SOAP::SOAPEnvelope.new(header, body) return send_soap(envelope, host) end # Builds and sends CreateAppointmentRequest to a provided Zimbra host. The # attachments argument expects a list of filename strings. # # Returns a SOAP::Mapping instance def self.createappointment(host, authToken, start, subject, description='', attachments=[]) header = SOAP::SOAPHeader.new context = element('context', nil, {'xmlns' => 'urn:zimbra'}, [ element('authToken', authToken) ]) header.add('context', context) aids = [] for attachment in attachments aids << upload_attachment(host, authToken, attachment) end if aids.empty? attach = nil else attach = element('attach', nil, { 'aid' => aids.join(",") }) end body = SOAP::SOAPBody.new(element('CreateAppointmentRequest', nil, { 'xmlns' => 'urn:zimbraMail' }, [ element('m', nil, {}, [ element('inv', nil, {}, [ element('comp', nil, { 'status' => 'CONF', 'allDay' => 1, 'fb' => 'F', 'name' => subject, 'noBlob' => 1, }, [ datetime('s', start), datetime('e', start), element('descHtml', description), element('alarm', nil, { 'action' => 'DISPLAY' }, [ element('trigger', nil, {}, [ element('rel', nil, { 'm' => 1 }) ]), element('desc', subject), ] ), ] ), ]), attach ]), ] )) envelope = SOAP::SOAPEnvelope.new(header, body) send_soap(envelope, host) end # builds SOAP::SOAPElement with tag name with a *d* attribute of the # provided ruby Time def self.datetime(name, time) return element(name, nil, {'d' => time.strftime("%Y%m%d")}) end # builds SOAP::SOAPElements the way SOAP::SOAPElement constructor _should_ # # element('AuthRequest', nil, # { # 'xmlns' => 'urn:zimbraAdmin', # }, # [ # element('name', 'whoa'), # element('password', 'man'), # ] # ) # # The returned SOAP::SOAPElement converted to XML would be: # # <AuthRequest xmlns="urn:zimbraAdmin"> # <name>whoa</name> # <password>man</password> # </AuthRequest> def self.element(name, value=nil, attrs={}, children=[]) element = SOAP::SOAPElement.new(name, value) element.extraattr.update(attrs) for child in children if child element.add(child) end end return element end # Builds and sends DelegateAuth Request to a provided Zimbra host. The # authToken must be that of an admin! The account arg is nothing fancy, just # the username of the user to spoof. # # Returns a SOAP::Mapping instance, with an authToken attribute def self.masquerade(host, authToken, account) header = SOAP::SOAPHeader.new context = element('context', nil, {'xmlns' => 'urn:zimbra'}, [ element('authToken', authToken) ]) header.add('context', context) body = SOAP::SOAPBody.new(element('DelegateAuthRequest', nil, { 'xmlns' => 'urn:zimbraAdmin' }, [ element('account', account, { 'by' => 'name', }) ] )) envelope = SOAP::SOAPEnvelope.new(header, body) return send_soap(envelope, host) end # Marshals SOAP::Envelopes and sends them to a given Zimbra host # # Returns response as a SOAP::Mapping instance def self.send_soap(envelope, host) url = 'https://' + host + ':7071/service/admin/soap/' stream = SOAP::HTTPStreamHandler.new(SOAP::Property.new) request_string = SOAP::Processor.marshal(envelope) puts request_string if $DEBUG request = SOAP::StreamHandler::ConnectionData.new(request_string) response_string = stream.send(url, request).receive_string puts response_string if $DEBUG env = SOAP::Processor.unmarshal(response_string) return SOAP::Mapping.soap2obj(env.body.root_node) end # Uploads file to given Zimbra host # # Returns a string containing the Zimbra attachment id. These attachments are # only accessible to the user that uploaded them. def self.upload_attachment(host, authToken, filename) params = Hash.new file = File.open(filename, "rb") params["attachment"] = file mp = Multipart::MultipartPost.new query, headers = mp.prepare_query(params) file.close headers['Cookie'] = 'ZM_AUTH_TOKEN=' + authToken url = URI.parse('https://' + host + '/service/upload') client = Net::HTTP.new(url.host, url.port) client.use_ssl = true response = client.post(url.path + '?fmt=raw', query, headers) return response.body.split(',')[2].strip.slice(1..-2) end end
Note: I would have done this in python, if it were not needed for an existing rails application. ;)
I found a blog post today that gleans the names and messages from the Twitter search. As an exercise, I decided to rewrite this using mechanize and lxml. My code writes to the standard out instead of a file. The user could redirect the output for the same effect. Note: I am aware that Twitter has JSON, plus several apis, and using that would be easier than this. This is an exercise.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 | #!/usr/bin/env python import getopt import sys from mechanize import Browser, _mechanize from lxml.html import parse baseurl = "http://search.twitter.com/search?lang=en&q=" def search_twitter(terms, pages=1): """ terms = a list of search terms pages(optional) = number of pages to retrive returns a list of dictionaries """ br = Browser() br.set_handle_robots(False) results = [] response = br.open("".join([baseurl, "+".join(terms)])) while(pages > 0): doc = parse(response).getroot() for msg in doc.cssselect('div.msg'): name = msg.cssselect('a')[0].text_content() text = msg.cssselect('span')[0].text_content() text = text.replace(' (expand)', '') results.append({ 'name': name, 'text': text, }) try: response = br.follow_link(text='Older') except _mechanize.LinkNotFoundError: break # No more pages :( pages -= 1 return results if __name__ == '__main__': optlist, args = getopt.getopt(sys.argv[1:], 'p:', ['pages=']) optd = dict(optlist) pages = 1 if '-p' in optd: pages = int(optd['-p']) if '--pages' in optd: pages = int(optd['--pages']) if len(args) < 1: print """ Usage: %s [-p] [--pages] search terms -p, --pages = number of pages to retrieve """ % sys.argv[0] raise SystemExit, 1 results = search_twitter(args, pages) for result in results: print "%(name)-20s%(text)s" % result |
In my post from a while back, I gave an example of the standard HTMLParser's use. HTMLParser is not the easiest way to glean information from HTML. There are two modules that are not part of the standard python distribution that can shorten the development time. The first is BeautifulSoup. Here is the code from the previous episode using BeautifulSoup instead of HTMLParser.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 | #!/usr/bin/env python from BeautifulSoup import BeautifulSoup from mechanize import Browser br = Browser() response = br.open('http://tylerlesmann.com/') soup = BeautifulSoup(response.read()) headers = soup.findAll('div', attrs={'class': 'header'}) headlines = [] for header in headers: links = header.findAll('a') for link in links: headlines.append(link.string) for headline in headlines: print headline |
This is a lot shorter, 16 instead of 38 lines. It also took about 20 seconds to write. There is one gotcha here. Both scripts do the same task. This one using BeautifulSoup takes over twice as long to run. CPU time is much cheaper than development time though.
The next module is lxml. Here's the lxml version of the code.
1 2 3 4 5 6 7 8 9 10 | #!/usr/bin/env python from lxml.html import parse from mechanize import Browser br = Browser() response = br.open('http://tylerlesmann.com/') doc = parse(response).getroot() for link in doc.cssselect('div.header a'): print link.text_content() |
As you can see, it is even shorter than BeautifulSoup at 10 lines. On top of that, lxml is faster than HTMLParser. So what is the catch? The lxml module uses C code, so you will not be able to use it on Google's AppEngine or on Jython.
In the last post, I illustrated how to most efficiently fetch html for data mining using the mechanize module. Now that we have our html, we can parse it for the information we want. To do this, we will use the HTMLParser module. This is a standard module in Python, so you don't have to install anything.
In this example, we will glean all of the headlines from the main page of this blog.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 | #!/usr/bin/env python from HTMLParser import HTMLParser from mechanize import Browser class HeadLineParser(HTMLParser): def __init__(self): self.in_header = False self.in_headline = False self.headlines = [] HTMLParser.__init__(self) def handle_starttag(self, tag, attrs): if tag == 'div': # attrs is a list of tuple pairs, a dictionary is more useful dattrs = dict(attrs) if 'class' in dattrs and dattrs['class'] == 'header': self.in_header = True if tag == 'a' and self.in_header: self.in_headline = True def handle_endtag(self, tag): if tag == 'div': self.in_header = False if tag == 'a': self.in_headline = False def handle_data(self, data): if self.in_headline: self.headlines.append(data) br = Browser() response = br.open('http://tylerlesmann.com/') hlp = HeadLineParser() hlp.feed(response.read()) for headline in hlp.headlines: print headline hlp.close() |
You use HTMLParser by extending it. The four functions you'll need are init, handle_starttag, handle_endtag, and handle_data. HTMLParser can be confusing at first because it works in a unique matter. Whenever a html tag is encountered, handle_starttag is called. Whenever a closing tag is found, handle_endtag is called. Whenever anything in between tags is encountered, handle_data is called.
The way to actually use HTMLParser is to use a system of flags, like in_header and in_headline from the example. We toggle them on in handle_starttag and off in handle_endtag. If you look at the html of this blog, you'll see that headlines are enclosed in classless <a> tags. There are alot of <a>s on this site. We need something unique to flag the headline <a>s. If you look carefully, you would see that all of the headlines are enclosed with <div>s with a header class. We can flag those and flag the <a>s only inside them, which is what the example does.
Now that the script has all the proper flags to detect headlines, we can simply have handle_data append any text to a list of headlines when our in in_headline flag is True.
To use our new parser, we simply make an instance of it and use the instance's feed method to run html through the parser. We can access the headlines attribute directly like we can in any object in python.
