Cacti, monitoring, SNMP, Linux and OSX

9 February 2007

I setup Cacti to easily monitor my Linux PC and my PowerBook via SNMP.
Here are a few notes about the process :

  • install Cacti : yum install cacti.
  • install the dependencies : net-snmp-utils. Where is cactid ??
  • create the database, create the mysql user : cactiuser for example (do not confuse with the cacti Linux user !), and initialise the database /usr/share/doc/cacti-0.8.6.i/cacti.sql
  • Configure the /etc/cacti/db.php accordingly.
  • Enable the polling in cron : /etc/cron.d/cacti
  • Eventually authorize more IP in /etc/httpd/conf.d/cacti.conf
  • Open http://localhost/cacti/ in a browser and add devices, graphs …

Compared to some other sources, as of Fedora core 6 (FC6), I didn’t have to change user right on directories. I don’t know about SELinux, I should test.

On the PowerBook Mac OS X : set up net-snmp (General SNMP Cacti’s howto) :

  • sudo vi /usr/share/snmp/snmpd.conf. Here is the simplest configuration I could do, still providing SNMPv3 security.

    createUser cacti MD5 password DES
    rouser cacti
    includeAllDisks 5%
    syslocation On my desk
    syscontact toto.tata@gmail.com

  • Configure the SNMPv3 user/password in Cacti too (Menu : “Settings”) !
  • Stop snmpd : sudo killall snmpd
  • Restart snmpd : sudo snmpd -I -icmp. See : Net-SNMP README OSX.
  • You can have a look at the log file : /var/log/snmpd.log. I still have some errors, but it works :

    nlist err: neither boottime nor _boottime found.
    nlist err: neither cnt nor _cnt found.
    kvm_read(*, 0, 0x249eec, 4) = 0: kvm_read: Bad address
    auto_nlist failed on boottime at location 0


Yahoo pipes

8 February 2007

This is really great : try it !! For example, I made a Search limited to your delicious bookmarked sites.
OK It is a BETA : a bit unstable or maybe it is my browser and a bit slow and not perfect yet, errors doesn’t seems to be reported, but once it will be reliable, this will be great.
(Looks like it is very server side for now. So maybe they have to much success.)

At least, that kind of tool may drive the adoption of more “processing friendly” techno, like XHTML and other Semantic-web stuff that only geek seems to care about now.

Via Yahoo! Pipes: Unlocking the Data Web. Read it, it gives much more info.