Host Configuration File Examples
This section covers various configuration examples and how they affect the host configuration files. The first host configuration file will be shown server00.yml, and the host configuration files for the remaining servers will follow the same pattern. For more complex examples, two host configuration files will be shown to help better understand the configuration.
Example 1: Data Network Interface Manually Configured
Host configuration with no data interfaces (data_ifaces) provided. The playbook does not need to configure the IP on the data interface. The user configured the interfaces prior to running the playbook.
---nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount6 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)Example 2: Data Network Interface Automatically Configured
Host configuration with a single data interface. The playbook configured the interface. The user did not configure interface with the data IP.
---data_ifacesbootprotostatic conn_nameens1 ifnameens1 ip410.10.10.100/24nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount6 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)The example above shows the format for setting IPv4 addresses using ip4: ip/subnet. IPv6 addresses can be set using ip6: ip/prefix.
Example 3: Override the Lightbits Configurations
Host configuration with Lightbits override. The provided value overrides the key listen-client-urls.
---nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount6 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)etcd_settings_user listen-client-urlshttp//127.0.0.12379,http//10.16.173.142379,http//192.168.16.2192379Example 4: Provide Custom Datapath Configuration
Host configuration with custom datapath configuration provided.
By default, the playbook inspects the remote machine and determines the directory containing the specific configuration for Duroslight and backend services (datapath configuration). The excluding node-manager configuration uses the following logic:
<system_vendor>-<processor_count>-processor-<processor_cores>-cores---nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount6 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)datapath_configcustom-datapath-configExample 5: Use the Linux Volume Manager (LVM) Partition for etcd Data
Host configuration with custom lvm partition for etcd data.
---nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount6 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)use_lvm_for_etcdtrueetcd_lv_nameetcd#etcd_settings_user:etcd_lv_size15GiBetcd_vg_nameAlmaExample 6: Profile-Generator Overrides
Enable humans to override profile-generator output and provide for each server a custom file that will be taken by profile-generator as the system-profile.
Each host can be different so each host can specify its own override file.
nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue storageDeviceLayout initialDeviceCount6 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatcherspartition == falsesize >= gib(300)profile_generator_overrides_dir/tmp/overrides.d/server00In case the cluster is homogeneous and we want to apply the same override to all nodes, we can provide a single setting in the groups/all.yml file
or via the cmd with:
ansible-playbook -i ansible/inventories/cluster_example/hosts playbooks/deploy-lightos.yml -e profile_generator_overrides_dir=/tmp/overrides.dIn the above example, we specify profile_generator_overrides_dir, which is a directory on the Ansible Controller that will be copied to the target machine.
Example 7: Dual Instance Configuration
In a dual instance configuration. Lightbits is run from both NUMAs of the CPU.
Below is an example configuration for server00.yml and server01.yml.
In this configuration, both servers NUMA0 and 1 have 12 NVME drives, for a total of 24 NVME drives in the server.
Each server has a single managment IP configured (not shown below) and two data IPs (one for each NUMA). Each instance has a unique data IP set for it.
server00.yml
---nameserver00nodesinstanceID0 data_ip172.16.10.10 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount12 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)# - name =~ "nvme0n1"instanceID1 data_ip172.16.20.10 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount12 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)server01.yml
---nameserver01nodesinstanceID0 data_ip172.16.10.11 failure_domainsserver01 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount12 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)# - name =~ "nvme0n1"instanceID1 data_ip172.16.20.11 failure_domainsserver01 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount12 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)Example 8: Single IP Dual NUMA Configuration
This example is for Single IP Dual Instance configuration. For more information, see Single-IP-Dual-NUMA Configuration.
The below is an example of a full server00.yml, with a similar physical configuration as the above example. However, instead of using two data IPs, it only uses one data IP for both of its instances.
Unlike dual instance configuration from the example above, which had a unique data IP per instance, each instance on a server has the same IP.
---nameserver00nodesinstanceID0 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount12 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)# - name =~ "nvme0n1"instanceID1 data_ip10.10.10.100 failure_domainsserver00 ec_enabledtrue lightfieldModeSW_LF storageDeviceLayout initialDeviceCount12 maxDeviceCount12 allowCrossNumaDevicesfalse deviceMatchers# - model =~ ".*"partition == falsesize >= gib(300)