LISA Conference 2002 Notes

http://www.usenix.org/lisa02/

Tutorial Programs

• Sunday November 3, 2002
• Blueprints for High Availability: Designing Resilient Distributed Systems
  • Evan Marcus, Veritas Software Corp. evan@veritasPLEASENOSPAM.com
  • Drop Evan an email to get updated slides
  • "High Availability cannot be achieved by merely installing Failover Software and walking away."
  • Keep Availability Statistics
    • Include root cause of problems if possible
  • System naming conventions
    • Use mnemoic names - violate rules of simplicity
    • Add personification
  • Security Book
    • Secret's & Lies - Bruce Schnier
  • Crack to check passwords

• Monday November 4, 2002
  • Exploring the Potential of LDAP
    • Gerald Carter, jerry@sambaPLEASENOSPAM.org
    • OpenLDAP http://www.openldap.org
    • SunOne Directory Server http://wwws.sun.com/software/product_categories/directory_servers_identity_mgmt.html
    • Novell's NDS eDirectory 8.x http://www.novell.com/products/edirectory
    • Always be ready to throw away your first implimentation, even your second. * You will not get it right the first time
    • dc=principal,dc=com
    • Directory Mark http://www.mindcraft.com/directorymark
      • Benchmarking
    • Finding out what indexes for openldap
      • -256 log level to see what search string for indexes
      • ethereal - packet sniffer that works good for sniffing openldap to see what is being asked for
    • Put /etc/services in LDAP?
    • ou=people ou=group ou=hosts in RFC 2307biz
    • Check http://www.plainjoe.org/ldap/scripts/ for configure.tar.gz for config examples used in class
    • Start_TLS highly recommended or else passing plain text
    • Devices LDAP server, Users LDAP server
    • http://www.louisville.edu/~tjande01/openldap/ OpenLDAP schema for AIX

• Tuesday November 5, 2002
  • Intro to Massive Upgrades and Changes
    • Christine Hogan and Tom Limoncelli
    • "The Practice of System and Network Administration" by Limoncelli and Hogan
    • http://www.EverythingSysadmin.com
    • Create a list of services, software, users on a checklist
    • netstat -a, rc files, ps
    • Backout plan
    • Notification communicated
    • Test verifications
    • Upgrade
    • Test verifications, debug
    • Notify upgrade completed to same community

• • Advanced Console Remote Access
    • Notes for session are very good and in the book
    • http://www.americable.com
    • LDAP authentication for Cyclades new box...Does it work with the older hardware?
    • http://www.conserver.com/consoles/

Technical Sessions

• Wednesday November 6, 2002
  • AFS Guru session
    • Not very good, doesn't really seam like something that would work easily for us.
  • The Constitutional and Financial Arguement Against SPAM.
    • Dan Klein, dvk@lonewolfPLEASENOSPAM.com
    • http://wwww.klein.com for presentation slides with audio
• Thursday November 7, 2002
  • Monitoring and Logging
    • Addamark search
      • Used for large search engines, yahoo, Atom, etc.
    • Mielog - Visual Log Browser
    • Detecting Events that Didn't happen
      • Something that should usually have happened that didn't
    • Defining and Monitoring Service Level Agreements for dynamic e-Business
      • Why do sysadmins care about SLAs?
      • How much does it cost to guarantee a Response time less than 1 sec?
      • 3rd party for web service monitoring
    • Hotswap - Transparent Server Failover for Linux
      • Maintain client connections by using own IP layer to make connections to two machines
    • Defending Against Internet Attacks - SOrting Through the Value Propositions of Different Security Technologies
      • Paul E. Proctor, Practical Security, Inc.
      • notes and slides on http://www.practicalsecurity.com

• Friday November 8, 2002
  • • Performance Tuning Guru
      • Jeff Allen, Tellme Networks
      • VXFS, fsadm checking for fragmenting, how much is too much? Increased kernel versus user time.
      • ncsize, kernel parameter in Solaris, name cache size avoid the work of looking up inodes.
        • explicity set ncsize in /etc/system dynamically kernel size adjustments. High kernel time due to large "vmstat -s" dnlc hit rate should be over 90%. Think about application why would it be low.
      • general approach for finding problems
        • Scientific point of view...measure first, put in new setting, measure again.
      • Where is my bottleneck? I/O bound?(iostat -x) whatching on a per controller basis, then disk basis. Watch right hand column. %busy 40-55% is OK. Memory bound, SR Scan Rate in vmstat.
      • Logging file systems could help mail system by allowing the whole process happen in the log. File creating and deleting could happen all in the log.
  • System Monitoring Guru Session
    • Doug Hughes, monitoring guru
      • Netcool - http://www.micromuse.com/products/netcool_suite_overview.html - Premiere alert, event coralation tool. GUI for prioritizing.
        • Object Server, database underneath, deduplication, correlation, automation, resolution
        • Probes, active or passive
        • 30,000 events/minute
        • Down and up event coralation
        • Syslog or snmp probes are most popular
        • VERY $$$$$$$$$$$$$$ $4million to $100,000
        • very good customer support
        • Dynamic suppression
      • Nagios http://www.nagios.com
        • is a host and service monitor designed to inform you of network problems before your clients, end-users or managers do.
        • Static suppression
  • Infrastructures Guru Session
    • Steve Traugott, http://www.infrastructures.org
      • Pull works best because some machine will be down.
      • Pull helps divergence. Host brings itself up to the latest rev.
      • Push is controlled, infrastructure is not overwhelmed.
      • Do both pushes and pull ( Pull am I up to date ) ( Trigger pulls )

To Do

• Crack for passwords
• Download OpenLDAP, get something running
  • OpenLDAP running on geeko now. Need to work on getting acount into the directory for authentication testing.
• Check out I/O on mail systems, do we have logging turned on?
• Evaluate
  • Netcool http://www.micromuse.com/products/netcool_suite_overview.html - Network monitoring
  • Nagios http://www.nagios.com - Network monitoring
  • Team Quest http://www.teamquest.com - Capacity Planning and Performance Monitoring

-- MattMillard - 09 Nov 2002