2
0
mirror of https://github.com/xcat2/xcat-core.git synced 2024-11-24 10:30:01 +00:00
16 DNS_master_slave_support
ligc edited this page 2015-07-30 04:22:49 -04:00

Table of Contents

Design_Warning

Background

This requirement is from the scenario where the users want to use xcat SN as DNS failover, but current xcat DNS structure can not support it because current SN acts as DNS forwarding/caching server, so when MN goes down, SN can not be working independently. See more details in the following defect: https://sourceforge.net/p/xcat/bugs/3863/

Basic Idea

Using DDNS master/slave configuration can meet this requirement, after making corresponding configuration on DNS master/slave(/etc/named.conf), the master allows to transfer DNS zones to the DNS slaves, it means when DNS records are updated dynamically(for example, makedns is issued), the auto-sync will be done between the master and slaves without xcat's involvement. So this can make sure DNS records synced between MN and SNs, even MN goes down, SNs can be working for name resolution.

External User Interface

  • add to the servicenode.nameserver attribute an additional valid value, 2, that means make it a slave dns (a value of 1 will still mean a forwarding dns).
  • makedsn -n can check to see if there are any SNs defined in the servicenode table with nameserver=2, and if so, add them to MN's named.conf.
  • if they add their SN definitions in the db after they initially ran makedns -n, then they need to run makedns -n again after adding the SNs.
  • for the config of named.conf on the SN, it can still be done by AAsn.pm when xcatd starts on the SN.

This setting(servicenode.nameserver=2) only impacts the behavior of "makedns -n" to generate a different /etc/named.conf supporting dns master/slave mode, "makedns " or "makedns -d " can just keep the current logic to update DNS records on MN(DNS master), then named will be responsible to transfer/sync them to slaves.

Internal Implementation

ddns.pm

The code change in ddns.pm impacts /etc/named.conf on MN(dns master).

Modify the logic for /etc/named.conf generation, get servicenode.nameserver value, if it's 2, indicates servicenode.node(can be noderange or nodegroup) should be set as dns slave, generate /etc/named.conf as below (only list the difference here).

diff /etc/named.conf.master_slave /etc/named.conf.org

7,9d6 < allow-transfer {; ;...}; < notify yes; < also-notify {; ;...}; 45d41 <

=== AAsn.pm ===

The code change in AAsn.pm impacts /etc/named.conf on SNs(dns slaves), we need to add a new subroutine(for example make_named_conf_slave()) to generate /etc/named.conf with dns slave mode.

In subroutine setup_DNS(), get all the SNs with servicenode.nameserver=2, if I'm in the list(my hostname should be my xcat nodename as known by the management node), then call make_named_conf_slave(), otherwise, keep the current logic($XCATROOT/sbin/makenamed.conf).

The key points in /etc/named.conf on dns slaves are:

cat /etc/named.conf

options { allow-transfer {none;}; };

zone "xx.xx" in { type slave; masters {<master_IP>; }; file "db.xx.xx"; };

Documentation Change

Since there is user interface change for this implementation, so the documentation
[Cluster_Name_Resolution] needs to be updated accordingly.

Need to add comments in the Schema.pm file for the servicenode table ( nameserver attribute) for this new setting, so it will show up on man servicenode.

Other Design Considerations

  • '''Required reviewers''': xCAT ALL
  • '''Required approvers''': Bruce Potter
  • '''Database schema changes''': N/A
  • '''Affect on other components''': N/A
  • '''External interface changes, documentation, and usability issues''': Yes
  • '''Packaging, installation, dependencies''': N/A
  • '''Portability and platforms (HW/SW) supported''': N/A
  • '''Performance and scaling considerations''': N/A
  • '''Migration and coexistence''': Yes
  • '''Serviceability''': N/A
  • '''Security''': N/A
  • '''NLS and accessibility''': N/A
  • '''Invention protection''': N/A