Network Setup for Exadata Cloud Infrastructure Instances
This topic describes the recommended configuration for the VCN and several
related requirements for the Exadata Cloud Infrastructure
instance.
Before you set up an Exadata Cloud Infrastructure instance, you must set up a
virtual cloud network (VCN) and other Networking service
components.
VCN and Subnets To launch an Exadata Cloud Infrastructure instance, you must have a Virtual Cloud Network and at least two subnets:
Node Access to Object Storage: Static Route To be able to back up databases, and patch and update cloud tools on an Exadata Cloud Infrastructure instance, you must configure access to Oracle Cloud Infrastructure Object Storage.
Service Gateway for the VCN Your VCN needs access to both Object Storage for backups and Oracle YUM repos for OS updates.
To launch an Exadata Cloud Infrastructure
instance, you must have a Virtual Cloud Network and at least two subnets:
To launch an Exadata Cloud Infrastructure
instance, you must have a Virtual Cloud Network, at least two subnets and select the
type of DNS resolver you will use:
A VCN in the region where you
want the Exadata Cloud Infrastructure instance
At least two subnets in the VCN. The two subnets are:
Client subnet
Backup subnet
Choose which method of DNS name resolution you will use. See Choices for DNS in
Your VCN
The client subnet can be configured with either IPv4 or IPv4/IPv6 dual-stack networking.
In general, Oracle recommends using regional subnets , which span all
availability domains in the region. For more information, see Overview of VCNs and Subnets.
You will create custom route tables route tables for each subnet. You
will also create security rulessecurity rules to control traffic
to and from the client network and backup network of the Exadata compute nodes (for
The Cloud VM Cluster Resource, nodes are called virtual machines). More information follows about
those items.
Requirements for IP Address Space IP addresses must not overlap, especially when Exadata Cloud Infrastructure VM Clusters (and thus VCNs) are in more than one region.
DNS: Between On-Premises Network and VCN Oracle recommends using a private DNS resolver to enable the use of hostnames when on-premises hosts and VCN resources communicate with each other.
Option 1: Public Client Subnet with Internet Gateway π
This option can be useful when doing a proof-of-concept or development work.
You can use this setup in production if you want to use an internet gateway
with the VCN, or if you have services that run only on a public network and need
access to the database. See the following diagram and description.
Public client subnet (public means that the resources in the subnet can have public IP addresses at your discretion).
Private backup subnet (private means that the resources in the subnet cannot have public IP addresses and therefore cannot receive incoming connections from the internet).
Important See this known issue for information about configuring route rules with service gateway as the target on route tables associated with public subnets.
Custom route table for the private client subnet, with the following rules:
A rule for the on-premises network's CIDR, and target = DRG.
A rule for the service CIDR label called All <region> Services in Oracle Services Network, and target = the service gateway. The Oracle Services Network is a conceptual network in Oracle Cloud Infrastructure that is reserved for Oracle services. The rule enables the client subnet to reach the regional Oracle YUM repository for OS updates. Also see Option 2: Service Gateway Access to Both Object Storage and YUM Repos.
Optionally, a rule for 0.0.0.0/0, and target = NAT gateway.
Separate custom route table for the private backup subnet, with one rule:
The same rule as for the client subnet: for the service CIDR label called All <region> Services in Oracle Services Network, and target = the service gateway. This rule enables the backup subnet to reach the regional Object Storage for backups.
Optionally add a Static route on the compute
nodes to other OCI services (for VM clusters, the virtual machines) to enable
access, if the services are only reachable on the backup subnet and not via. the
client subnet, e.g. when using a NAT Gateway.
IP addresses must not overlap, especially when Exadata Cloud Infrastructure VM Clusters (and thus VCNs) are in more than one region.
If you're setting up Exadata Cloud Infrastructure VM Clusters (and thus VCNs) in more than one region, make sure the IP address space of the VCNs does not overlap. This is important if you want to set up disaster recovery with Oracle Data Guard.
For Exadata X8 and lower, the two subnets you create for the Exadata Cloud Infrastructure VM Clusters must not overlap with 192.168.128.0/20.
For Exadata X8M, X9M, and X11M, IP addresses (100.106.0.0/16 and 100.107.0.0/16) are used for the interconnect.
The following table lists the minimum required subnet sizes, depending on the Exadata VM Infrastructure for each VM Cluster size. For the client subnet, each node requires four IP addresses, and in addition, three addresses are reserved for Single Client Access Names (SCANs). For the backup subnet, each node requires three addresses.
Flexible infrastructure systems (X8M and higher) per VM Cluster
4 addresses per database node (minimum of 2 nodes) + 3
for SCANs
Minimum size determined by total number of IP addresses
needed for database nodes
3 address per database node (minimum of 2 nodes)
Minimum size determined by total number of IP addresses
needed for database nodes
Note
The Networking service reserves three IP addresses in each
subnet. Allocating a larger space for the subnet than the minimum
required (for example, at least /25 instead of /28) can reduce the relative impact
of those reserved addresses on the subnet's available space.
Configuring a Static Route for Accessing the Object Store π
All the traffic in an Exadata Cloud Infrastructure
instance is, by default, routed through the data network. To route backup traffic
to the backup interface (BONDETH1), you need to configure a static route on each
of the compute nodes in the cluster.
Setting Up DNS for an Exadata Cloud Infrastructure Instance
π
DNS lets you use host names instead of IP addresses to communicate with an Exadata
Cloud Infrastructure instance.
You can use the Internet and VCN Resolver (the DNS capability
built into the VCN) as described in DNS in Your Virtual Cloud
Network. Oracle recommends using a VCN Resolver for DNS name resolution for the
client subnet. It automatically resolves the Swift endpoints required for backing up
databases, patching, and updating the cloud tooling on an Exadata instance.
DNS: Short Names for the VCN, Subnets, and
Exadata Cloud Infrastructure instance
π
For the nodes to communicate, the VCN must use the Internet and VCN Resolver.
The Internet and VCN resolver enables hostname assignment to the nodes, and
DNS resolution of those hostnames by resources in the VCN.
The Internet and VCN resolver enables round robin resolution of the
database's SCANs. It also enables resolution
of important service endpoints required for backing up databases, patching, and updating
the cloud tooling on an Exadata Cloud Infrastructure
instance. The Internet and VCN Resolver is the VCN's default choice for DNS in the VCN.
For more information, see DNS in Your Virtual Cloud Network
and also DHCP Options.
When you create the VCN, subnets, and Exadata, you must carefully set the following identifiers, which are related to DNS in the VCN:
VCN domain label
Subnet domain label
Hostname prefix for the Exadata Cloud Infrastructure instance's cloud VM cluster or DB system resource
These values make up the node's fully qualified domain name (FQDN):
In this example, you assign exacs as the hostname prefix
when you create the cloud VM cluster or DB system. The Database service automatically
appends a hyphen and a five-letter string with the node number at the end. For
example:
Recommended maximum: 12 characters. For more information, see the example under the following section, "Requirements for the
VCN and subnet domain labels".
Cannot be the string localhost
Requirements for the VCN and subnet domain labels:
Recommended maximum: 14 characters each. The actual underlying requirement is a total of 28 characters across both domain labels (excluding the period between the labels). For example, both of these are acceptable: subnetad1.verylongvcnphx or verylongsubnetad1.vcnphx. For simplicity, the recommendation is 14 characters each.
No hyphens or underscores.
Recommended: include the region name in the VCN's domain label, and
include the availability domain name in the subnet's domain label.
In general, the FQDN has a maximum total limit of 63 characters. Here is a safe
general rule:
Private view and private zone must be created before launching DB system provisioning. For details, see Private DNS resolvers.
Forwarding to another DNS server should be set up beforehand in the DNS console. This can be done by going to the VCN's resolver, and creating the endpoint and then the rules. For details, see DNS in Your Virtual Cloud Network.
Private zone's name cannot have more than 4 labels. For example, a.b.c.d is allowed while a.b.c.d.e is not.
When provisioning a Exadata VM Cluster using Private DNS feature, Exadata needs to create reverse DNS zones in the compartment of Exadata VM Cluster. If the compartment has defined tags or tag defaults, additional policies related to managing tags are needed. For details, see:
To be able to back up databases, and patch and update cloud tools on an Exadata Cloud Infrastructure instance, you must configure
access to Oracle Cloud Infrastructure Object Storage. Regardless of how you
configure the VCN with that access (for example, with a service gateway), you may also need
to configure a static route to Object Storage on each of the compute nodes in the cluster.
This is only required if you are not using automatic backups. If you are using customized
backups using the backup APIs, then you must route traffic destined for Object Storage
through the backup interface (BONDETH1). This is not necessary if you are using the
automatic backups created with the Console, APIs, or CLIs.
Caution:
You must configure a
static route for Object Storage access on each compute node in an Exadata Cloud Infrastructure instance if you are
not creating automatic backups with the Console, APIs, or CLIs. Otherwise,
attempts to back up databases, and patch or update tools on the system, can fail.
Note
When you enable the first automatic backup for a database the static route
configuration will be automatically done on the service.
If you want to patch the service before creating a database, the manual static route
is required to be able to patch the GI or DB Home.
The static route may also be required to access other services (IAM, KMS) if these
are not reachable via client subnet and only the backup subnet uses the setting to
access all servcies within a region.
Oracle Cloud Infrastructure Object Storage uses the CIDR block IP range
134.70.0.0/16 for all regions.
As of June 1, 2018, Object Storage no longer supports the following
discontinued IP ranges. Oracle recommends that you remove these older IP addresses from
your access-control lists, firewall rules, and other rules after you have adopted the
new IP ranges.
Your VCN needs access to both Object Storage for backups and Oracle YUM repos for OS
updates.
Depending on whether you use Option1: Public Client Subnet with Internet
Gateway or Option 2: Private Subnets described previously, you use the service gateway
in different ways. See the next two sections.
Option 1: Service Gateway Access to OCI
Services π
You configure the backup subnet to use the service gateway for access only to
Object Storage.
As a reminder, here's the diagram for option 1:
In general, you must:
Perform the tasks for setting up a service gateway on a VCN, and
specifically enable the service CIDR label called OCI <region> Object Storage.
In the task for updating routing, add a route rule to the backup subnet's custom route table. For the destination service, use OCI <region> Object Storage and target = the service gateway.
In the task for updating security rules in the subnet, perform the task on the
backup network's network security group (NSG) or custom security list.
Set up a security rule with the destination service set to OCI <region> Object Storage. See Rule
Required Specifically for the Backup NetworkRule Required Specifically for the Backup
Network .
Option 2: Service Gateway Access to Both Object Storage and YUM Repos π
You configure both the client subnet and backup subnet to use the service
gateway for access to the Oracle Services Network, which includes both Object Storage and
the Oracle YUM repos.
Note
See this known issues for information
about accessing Oracle YUM services through the service gateway.
As a reminder, here's the diagram for option 2:
In general, you must:
Perform the tasks for setting up a service gateway on a VCN, and
specifically enable the service CIDR label called All <region> Services in Oracle Services Network.
In the task for updating routing in each subnet, add a rule to each subnet's custom route table. For the destination service, use All <region> Services in Oracle Services Network and target = the service gateway.
In the task for updating security rules for the subnet, perform the task on the
backup network's network security group (NSG) or custom security list.
Set up a security rule with the destination service set to OCI <region> Object Storage. See Security
Rules for the Oracle Exadata Cloud ServiceSecurity Rules for the Oracle Exadata Database Service on Dedicated Infrastructure. Note that the client subnet already has a broad egress rule that
covers access to the YUM repos.
Here are a few additional details about using the service gateway for option 2:
Both the client subnet and backup subnet use the service gateway, but to access different services. You cannot enable both the OCI <region> Object Storage service CIDR label and the All <region> Services in Oracle Services Network for the service gateway. To cover the needs of both subnets, you must enable All <region> Services in Oracle Services Network for the service gateway. The VCN can have only a single service gateway.
Any route rule that targets a given service gateway must use an enabled service CIDR label and not a CIDR block as the destination for the rule. That means for option 2, the route tables for both subnets must use All <region> Services in Oracle Services Network for their service gateway rules.
Unlike route rules, security rules can use either any service CIDR label
(whether the VCN has a service gateway or not) or a CIDR block as the source or
destination CIDR for the rule. Therefore, although the backup subnet has a route
rule that uses All <region> Services in
Oracle Services Network, the subnet can have a security rule that uses
OCI <region> Object Storage. See
Security Rules for the Exadata Cloud Service instance.
Security Rules for the Oracle Exadata Database Service on Dedicated
Infrastructure π
This section lists the security rules to use with Exadata Cloud Infrastructure.
Security rules control the types of traffic allowed for the client network and
backup network of the Exadata's compute nodes. The
rules are divided into three sections.
For X8M and X9M systems, Oracle
recommends that all ports on the client subnet need to be open for ingress and
egress traffic. This is a requirement for adding additional database servers to the
system.
Rules Required for Both the Client Network and Backup Network
This section has several general rules that enable essential connectivity for hosts in the VCN.
If you use security lists to implement your security
rules, be aware that the rules that follow are included by default in the default security list. Update
or replace the list to meet your particular security needs. The two ICMP rules
(general ingress rules 2 and 3) are required for proper functioning of network
traffic within the Oracle Cloud Infrastructure environment. Adjust the
general ingress rule 1 (the SSH rule) and the general egress rule 1 to allow traffic
only to and from hosts that require communication with resources in your VCN.
General ingress rule 1: Allows SSH traffic from anywhere
Stateless: No (all rules must be stateful)
Source Type: CIDR
Source CIDR: 0.0.0.0/0 (IPv4), ::/0 (IPv6)
IP Protocol: SSH
Source Port Range: All
Destination Port Range: 22
General ingress rule 2: Allows Path MTU Discovery fragmentation messages
This rule enables hosts in the VCN to receive Path MTU Discovery fragmentation messages. Without access to these messages, hosts in the VCN can have problems communicating with hosts outside the VCN.
Stateless: No (all rules must be stateful)
Source Type: CIDR
Source CIDR: 0.0.0.0/0 (IPv4), ::/0 (IPv6)
IP Protocol: ICMP
Type: 3
Code: 4
General ingress rule 3: Allows connectivity error messages within the VCN
This rule enables the hosts in the VCN to receive connectivity error messages from each other.
Stateless: No (all rules must be stateful)
Source Type: CIDR
Source CIDR: IPv4: your VCN's IPv4 CIDR, IPv6: your VCN's IPv6 CIDR
IP Protocol: ICMP
Type: ALL
Code: All
General egress rule 1: Allows all egress traffic
Stateless: No (all rules must be stateful)
Destination Type: CIDR
Destination CIDR: 0.0.0.0/0 (IPv4), ::/0 (IPv6)
IP Protocol: All
Rules Required Specifically for the Client Network
The following security rules are important for the client network.
Note
Client ingress rules 1 and 2 only cover connections initiated from within the client subnet. If you have a client that resides outside the VCN, Oracle recommends setting up two additional similar rules that instead have the Source CIDR set to the public IP address of the client.
Client egress rules 1 and 2 in the client network configuration allow TCP and ICMP traffic, enabling secure node-to-node communication within the client network. These rules are critical, as they facilitate essential TCP connectivity across nodes. Should TCP connectivity fail between nodes, provisioning of the Exadata Cloud VM cluster resource will not complete successfully.
Client ingress rule 1: Allows ONS and FAN traffic from within the client subnet
The first rule is recommended and enables the Oracle Notification Services (ONS) to communicate about Fast Application Notification (FAN) events.
Client egress rule 2: Allows all egress traffic (allows connections to the Oracle YUM repos)
Client egress rule 2 is important because it allows connections to the Oracle YUM repositories. It is redundant with the general egress rule in this topic (and in the default security list). It is optional but recommended in case the general egress rule (or default security list) is inadvertently changed.
Stateless: No (all rules must be stateful)
Destination Type: CIDR
Destination CIDR: 0.0.0.0/0 (IPv4), ::/0 (IPv6)
IP Protocol: All
Description: An optional description of the rule.
Rule Required Specifically for the Backup Network
The following security rule is important for the backup network because
it enables the DB system to communicate with Object Storage through the service
gateway (and optionally with the Oracle YUM repos if the client network doesn't have
access to them). It is redundant with the general egress rule in this topic (and in
the default security list). It is
optional but recommended in case the general egress rule (or default security list)
is inadvertently changed.
Backup egress rule: Allows access to Object Storage
Stateless: No (all rules must be stateful)
Destination Type: Service
Destination Service:
The service CIDR label called OCI <region> Object Storage
If the client network does not have access to the Oracle YUM repos, use the service CIDR label called All <region> Services in Oracle Services Network
Rule Required Specifically for the Backup Network The following security rule is important for the backup network because it enables the DB system to communicate with Object Storage through the service gateway (and optionally with the Oracle YUM repos if the client network doesn't have access to them).
Rules Required for Events Service The compute instance must have either a public IP address or a service gateway to be able to send compute instance metrics to the Events service.
Rules Required for Monitoring Service The compute instance must have either a public IP address or a service gateway to be able to send compute instance metrics to the Monitoring service.
Rules Required for Both the Client
Network and Backup Network π
This topic has several general rules that enable essential connectivity for
hosts in the VCN.
If you use security lists to implement your security rules, be aware that
the rules that follow are included by default in the default security list.
Update or replace the list to meet your particular security needs. The two ICMP rules
(general ingress rules 2 and 3) are required for proper functioning of network traffic
within the Oracle Cloud Infrastructure environment. Adjust the general ingress rule 1
(the SSH rule) and the general egress rule 1 to allow traffic only to and from hosts
that require communication with resources in your VCN.
General ingress rule 2: Allows Path MTU
Discovery fragmentation messages π
This rule enables hosts in the VCN to receive Path MTU Discovery
fragmentation messages. Without access to these messages, hosts in the VCN can have
problems communicating with hosts outside the VCN.
Rules Required Specifically for
the Client Network π
The following security rules are important for the client
network.
Note
For X8M systems, Oracle recommends that all ports on the client
subnet need to be open for ingress and egress traffic. This is a requirement for
adding additional database servers to the system.
Client ingress rules 1 and 2 only cover connections initiated from
within the client subnet. If you have a client that resides outside the
VCN, Oracle recommends setting up two additional similar rules
that instead have the Source CIDR set to the public IP address of the
client.
Client ingress rules 3 and 4 and client egress rules 1 and 2 allow
TCP and ICMP traffic inside the client network and enable the nodes to
communicate with each other. If TCP connectivity fails across the nodes, the
Exadata cloud VM cluster or DB system resource fails to provision.
Client egress rule 2: Allows all egress
traffic (allows connections to the Oracle YUM repos) π
Client egress rule 3 is important because it allows connections to the
Oracle YUM repos.
It is redundant with the general egress rule 1: Allow all egress traffic
(and in the default security list). It is optional but recommended in case the
general egress rule (or default security list) is inadvertently changed.
Rule Required Specifically for the Backup
Network π
The following security rule is important for the backup network because it
enables the DB system to communicate with Object Storage through the service gateway (and
optionally with the Oracle YUM repos if the client network doesn't have access to
them).
It is redundant with the general egress rule 1: Allows all egress
traffic in (and in the ). It is optional but recommended in case the general
egress rule (or default security list) is inadvertently changed.
The compute instance must have either a public IP address or a service
gateway to be able to send compute instance metrics to the Events service.
The default egress rules are sufficient to to allow the compute instance to
send compute instance metrics to the Events service.
If the instance does not have a public IP address, set up a service gateway
on the virtual cloud network (VCN). The service gateway lets the instance send compute
instance metrics to the Events service without the traffic going over the internet. Here
are special notes for setting up the service gateway to access the Events service:
When creating the service gateway, enable the service label called
All <region> Services in Oracle Services Network. It includes the
Events service.
When setting up routing for the subnet that contains the instance,
set up a route rule with Target Type set to Service Gateway, and
the Destination Service set to All <region> Services in Oracle
Services Network.
The compute instance must have either a public IP address or a service
gateway to be able to send compute instance metrics to the Monitoring service.
The default egress rules are sufficient to to allow the compute instance to send compute
instance metrics to the Monitoring service.
If the instance does not have a public IP address, set up a service gateway on the
virtual cloud network (VCN). The service gateway lets the instance send compute instance
metrics to the Monitoring service without the traffic going over the internet. Here are
special notes for setting up the service gateway to access the Monitoring service:
When creating the service gateway, enable the service label called
All <region> Services in Oracle Services Network. It includes the
Monitoring service.
When setting up routing for the subnet that contains the instance,
set up a route rule with Target Type set to Service Gateway, and
the Destination Service set to All <region> Services in Oracle
Services Network.
When they choose the client subnet, they can also choose which
NSG or NSGs to use. Make sure they choose the client network's NSG.
When they choose the backup subnet, they can also choose which
NSG or NSGs to use. Make sure they choose the backup network's NSG.
You could instead create a separate NSG for the general rules. Then when the database administrator chooses which NSGs to use for the client network, make sure they choose both the general NSG and the client network NSG. Similarly for the backup network, they choose both the general NSG and the backup network NSG.
The new custom security list you created for the backup subnet.
Later when the database administrator creates the Exadata Cloud Service
instance, they must choose several networking components. When they select the client
subnet and backup subnet that you've already created and configured, the security rules
are automatically enforced for the nodes created in those subnets.
WARNING:
Do not remove the default egress rule from the default security list. If you do, make sure to instead include the following replacement egress rule in the client subnet's security list:
Network Requirements for Oracle
Database Autonomous Recovery Service π
Oracle Database Autonomous Recovery Service requires a registered
Recovery Service subnet dedicated to backup and recovery operations in your database virtual
cloud network (VCN).
To use Recovery Service for backups, follow the steps outlined in Configuring Recovery Service.
Create a Service Gateway to Object Storage In the OCI Console, create a service gateway to Object Storage. The service gateway is required for automation updates and configuration metadata.
In the OCI Console, create a service gateway to Object Storage. The
service gateway is required for automation updates and configuration metadata.
Open the navigation menu. Click Networking, and then
click Virtual Cloud Networks.
Select the VCN where your database services to be backed up are located.
On the resulting Virtual Cloud Network Details page, under
Resources,click Service
Gateways.
Click Create Service Gateway and provide the following
details.
Name: A descriptive name for the service gateway. It doesn't have
to be unique. Avoid entering confidential information.
Compartment: The compartment where you want to create the service
gateway, if different from the compartment you're currently working
in.
Services: Select the service CIDR Label, All
<region> Services in Oracle Services
Network from the drop-down list.
Tags: (advanced option) If you have permissions to create a
resource, then you also have permissions to apply free-form tags to that
resource. To apply a defined tag, you must have permissions to use the
tag namespace. For more information about tagging, see Resource
Tags. If you are not sure whether to apply tags, skip this
option (you can apply tags later) or ask your administrator.
Click Create Service Gateway.
Wait for the gateway to be created before proceeding to the next step.
Under Resources, click Route Tables.
Route Table Association: You can associate a specific VCN route table
with this gateway. If you associate a route table, afterward the gateway
must always have a route table associated with it. You can modify the rules
in the current route table or replace them with another route table.
Click the Route Table name that is being used by the
subnet for Recovery Service.
In the resulting Route Table Details page, click
Add Route Rules in the Route Rules
section.
When you configure a service gateway for a particular service CIDR label, you
must also create a route rule that specifies that label as the destination
and the target as the service gateway. You do this for each subnet that
needs to access the gateway.
In the resulting Add Route Rules dialog, enter the
following details:
Target Type: Service Gateway.
Destination Service: The service CIDR label that is enabled for
the gateway. All <region> Services in
Oracle Services Network
Target Service Gateway: Select the name that you provided in step
4.