120 likes | 235 Vues
This experiment setup features Dell workstations running Redhat Linux connected to a DNS network for fault identification and symptom analysis. The architecture includes multiple master and slave DNS servers to test various issues and scenarios.
E N D
Autonomic DNS Experiment Architecture, Symptom and Fault Identification
Experiment Architecture • Physical system setup • Three Dell workstations running Redhat Linux 9.0, configured on an isolated network via IP Tables. • The network resides on the Computer Science Research network • Logical Domain Name System • Two Root servers controlling two top level domains: • .example • .test • Six sub-domains • red.test, yellow.test, green.test • white.example, orange.example, black.example
Experiment Architecture • All instances of the DNS will consist of Bind 9.2.3 • Each domain will consist of one master DNS. • Each domain will have 0 to 5 slave DNS. • Master (red) – ns.red.test • Slave (red) – ns.yellow.test, ns.green.test, ns.white.example, ns.orange.example, ns.black.example • Master (yellow) – ns.yellow.test • Slave (yellow) – ns.green.test, ns.white.example, ns.orange.example, ns.black.example From the examples above, each zone will have n-1 slave name servers assigned to it. The last name server will be without a slave.
Experiment Architecture • Having a varied number of slave name servers associated with the master name servers will allow us to test issues ranging from server performance on various levels to multiple user issues. • The experiments conducted will consist of the symptoms identified on the following slides
DNS Symptoms • Loss of Network Connectivity • Response from unexpected source • Recursion Bugs • Client unsure on handling of NS record in authority section • No answer to query • Client calls on server too many times • Name server is infected with bogus cache data
DNS Symptoms • A server refers to itself in the authority section • Cache leaks • Remote names can’t be looked up • Name error bugs • Lookups take a long time • Wrong or Inconsistent Answer • Slave name server data does not change when master server zone data changes • Is invalid proceeding anyway
DNS Symptoms • Slave server can’t load zone data • Internet services refused • Host fails authentication checks • Inconsistant or missing bad data • Lame server reported • Name server fails to load • Name server reports “Too many open files”
DNS Faults • Forgot to increment serial number • Forgot to reload primary master server after changes are made • Corrupt server cache • Ignored referral • To many referrals • Malicious server • Zero answer • Added name to db file, but forgot to add PTR record
DNS Faults • Name server cache set too small • Server does not do negative caching • Syntax error in zone data file on master • Incorrect IP address for master on slave zone data file • Syntax error in configuration file or zone data file • Missing dot at end of a domain name in zone data file
DNS Faults • Missing root.hints/db.cache data file • Missing subdomain delegation • TTL exceeded • Syntax error in resolv.conf • Incorrect labels in DNS name • Incorrect SOA format • Incorrect Glue records • Retry interval is set too low in SOA
DNS Faults • Incorrect address in query list – allow-query { address_match_list; }; • Incorrect configuration named.conf listen-on { ip_address; }; • PTR record points to CNAME • Expire time exceeded • Loss of network connectivity