feat(ec2): support multi NIC/IP setups #4799

aciba90 · 2024-01-24T11:21:48Z

Read individual commits for more context.

Additional Context

https://warthogs.atlassian.net/browse/SC-1662
https://bootstack.canonical.com/cases/00336928
https://bootstack.canonical.com/cases/00377150

Test Steps

Run the added integration tests on EC2.

Checklist

My code follows the process laid out in the documentation
I have updated or added any unit tests accordingly
I have updated or added any documentation accordingly

Merge type

Squash merge using "Proposed Commit Message"
Rebase and merge unique commits. Requires commit messages per-commit each referencing the pull request number (#<PR_NUM>)

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

…l#4799) Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 99-cloud-init-hook-hotplug.rules as ID_NET_DRIVER was not yet defined at matching-time. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

…l#4799) Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 99-cloud-init-hook-hotplug.rules as ID_NET_DRIVER was not yet defined at matching-time. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003

rpocase · 2024-01-30T12:43:51Z

I would be interested in @nmeyerhans 's feedback on this if you have some time. This is intended to provide multi-nic use cases for ubuntu based ec2 instances (i.e. a canonical supported alternative to ec2-net-utils). I don't know that this PR is intended to cover every use case yet.

TheRealFalcon

Overall, looks good! I left some comments inline.

More broadly, do we need to use separate routing tables to accomplish this? Netplan/V2 allow specifying routes per interface. Can specifying from on a route rather than specifying a table work for this?

If not, this code is netplan-specific (not v2-specific), and we'll need to gate it as such.

cloudinit/config/cc_install_hotplug.py

tests/integration_tests/modules/test_hotplug.py

tests/integration_tests/test_networking.py

cloudinit/sources/DataSourceEc2.py

nmeyerhans · 2024-01-31T04:02:51Z

More broadly, do we need to use separate routing tables to accomplish this? Netplan/V2 allow specifying routes per interface. Can specifying from on a route rather than specifying a table work for this?

The issue is that standard IP routing cannot take the source address of an outgoing packet into consideration. Source-based routing (aka policy routing) is necessary because VPCs (in the default configuration in which src/dest checking is enabled) enforce that packets egressing an ENI must have a source address that matches the list of addresses or prefixes assigned to that ENI in the VPC control plane. Packets that do not meet these criteria are dropped. Policy routing, which allows routing decisions based on additional factors such as the source address, is necessary in order to comply with VPCs restrictions.

A common scenario that triggers this relates to return traffic in response to an inbound packet to an instance with multiple attached ENIs. That traffic will come in with a destination IP matching one of the instance's interfaces. When the response is sent, that address will become the source address of the reply packet, and routing will be performed according to the entries in the main routing table. The routing decision will take the destination of the outgoing packet into account, but not the source. Thus it won't necessarily result in this response packet egressing the same ENI on which the original packet was received. If it egresses a different ENI, it will be dropped. The policy rules let us ensure that the traffic is routed according to a table that only knows about the interfaces with a given source address associated with them in the VPC control plane.

There isn't a way to ensure routing compliance with these VPC restrictions with standard destination-based routing behavior.

cloudinit/sources/DataSourceEc2.py

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

…l#4799) Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 99-cloud-init-hook-hotplug.rules as ID_NET_DRIVER was not yet defined at matching-time. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

…l#4799) Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 99-cloud-init-hook-hotplug.rules as ID_NET_DRIVER was not yet defined at matching-time. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

…l#4799) Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 99-cloud-init-hook-hotplug.rules as ID_NET_DRIVER was not yet defined at matching-time. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003

) Some instances, as p5 instances, can have multiple network cards and repeated device-numbers within them, see [0,1]. Add support to properly render the network configuration associated with them. References: [0] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#network-cards [1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/p5-efa.html

Some instances, as p5 instances, can have multiple network cards and repeated device-numbers within them, see [0,1]. Add support to properly render the network configuration associated with them. References: [0] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#network-cards [1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/p5-efa.html

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 90-cloud-init-hook-hotplug.rules as ID_NET_DRIVER is not defined until [2]80-net-setup-link.rules is sourced. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003 [2] https://github.com/systemd/systemd/blob/main/rules.d/80-net-setup-link.rules

Some instances, as p5 instances, can have multiple network cards and repeated device-numbers within them, see [0,1]. Add support to properly render the network configuration associated with them. References: [0] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#network-cards [1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/p5-efa.html

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

Avoids LP: #1946003 on upgraded systems. References: [0] #4799 [1] b519d86

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

Avoids LP: #1946003 on upgraded systems. References: [0] #4799 [1] b519d86

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 90-cloud-init-hook-hotplug.rules as ID_NET_DRIVER is not defined until [2]80-net-setup-link.rules is sourced. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003 [2] https://github.com/systemd/systemd/blob/main/rules.d/80-net-setup-link.rules

Some instances, as p5 instances, can have multiple network cards and repeated device-numbers within them, see [0,1]. Add support to properly render the network configuration associated with them. References: [0] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#network-cards [1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/p5-efa.html

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

Avoids LP: #1946003 on upgraded systems. References: [0] #4799 [1] b519d86

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86 Co-authored-by: Chad Smith <chad.smith@canonical.com>

For EC2 instances with multiple NICs, policy-based routing will be configured on secondary NICs / secondary IPs to ensure outgoing packets are routed via the correct interface. Without this extra routing config, traffic coming via secondary NICs was routed using the main routing table, which can only contain one default route and the kernel only takes the destination IP address into account when selecting a route. Packets for destination beyond local networks were always routed through the default route, the one associated with the primary NIC. If traffic based on specific source IP addresses is associated with another NIC, wihtout these routing policies, this traffic would flow over the default route and the connection couldn't be established. References: [1] https://bootstack.canonical.com/cases/00336928 [2] https://bootstack.canonical.com/cases/00377150

Add extra logic to only trigger hook-hotplug on NICs with known drivers on EC2. This aviods the hook being triggered on any add/remove event on net devices, causing uneeded CPU usage, as on instance that start and stop a lot of docker containers, see [1]. Rename 10-cloud-init-hook-hotplug.rules to 90-cloud-init-hook-hotplug.rules as ID_NET_DRIVER is not defined until [2]80-net-setup-link.rules is sourced. References: [1] https://bugs.launchpad.net/cloud-init/+bug/1946003 [2] https://github.com/systemd/systemd/blob/main/rules.d/80-net-setup-link.rules

Some instances, as p5 instances, can have multiple network cards and repeated device-numbers within them, see [0,1]. Add support to properly render the network configuration associated with them. References: [0] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#network-cards [1] https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/p5-efa.html

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86 Co-authored-by: Chad Smith <chad.smith@canonical.com>

xdxu · 2024-04-29T00:27:48Z

This PR essentially broke the routing logic in our production code. We will have to clean up these redundant PBR rules to get things back to work.

IMHO, how we route the traffic (either local out or pass through traffic) in a multi-homed instance should be up to the end user, rather than relying on these automatically generated rules.

My comment is based on the assumption that Linux instances running on AWS EC2 is for generic purposes and we can't really assume that most users would like this new behavior. There are a lot of ways to implement multi-wan like routing scheme and most of them will be taking advantage of the PBR rules, which means these extra rules will likely break those systems.

aciba90 · 2024-04-29T14:21:08Z

@xdxu sorry about the breaking changes. We thought that providing a default working environment would add more value and that advanced users / use-cases could disable cloud-init's network configuration or provide the custom network configuration via cloud-init or other means.

If you still disagree, please open a new issue to discuss it with more visibility, or consider hanging out in #cloud-init IRC channel on Libera.

Thanks.

cloud-init (24.1.3-0ubuntu1~20.04.1) focal; urgency=medium * Upstream snapshot based on 24.1.3. (LP: #2056100). List of changes from upstream can be found at https://raw.githubusercontent.com/canonical/cloud-init/24.1.3/ChangeLog cloud-init (24.1.2-0ubuntu1~20.04.1) focal; urgency=medium * refresh patches: - d/p/retain-ec2-default-net-update-events.patch * Upstream snapshot based on 24.1.2. (LP: #2056100). List of changes from upstream can be found at https://raw.githubusercontent.com/canonical/cloud-init/24.1.2/ChangeLog cloud-init (24.1.1-0ubuntu1~20.04.1) focal; urgency=medium * d/apport-general-hook.py: Move apport hook to main branch * d/cloud-init.maintscript: remove /etc/cloud/clean.d/README * d/cloud-init.logrotate: add logrotate config for cloud-init * d/cloud-init.templates: enable WSL datasource by default * d/p/keep-dhclient-as-priority-client.patch: - Upstream switched to dhcpcd, keep isc-dhclient as the client * d/p/revert-551f560d-cloud-config-after-snap-seeding.patch - Retain systemd ordering cloud-config.service After=snapd.seeded.service * d/p/retain-ec2-default-net-update-events.patch: Reverts 4dbb08f5f0cc4f41cf9dd1474f0600a11510a3c9 to not change behavior on stable releases. * d/po/templates.pot: update for wsl * d/cloud-init.postinst: change priority of hotplug rules. Avoids LP #1946003 on upgraded systems. References: [0] canonical/cloud-init#4799 [1] commit/b519d861aff8b44a0610c176cb34adcbe28df144 * refresh patches: - d/p/netplan99-cannot-use-default.patch - d/p/retain-netplan-world-readable.patch - d/p/status-do-not-remove-duplicated-data.patch - d/p/status-retain-recoverable-error-exit-code.patch * Upstream snapshot based on 24.1.1. (LP: #2056100). List of changes from upstream can be found at https://raw.githubusercontent.com/canonical/cloud-init/24.1.1/ChangeLog

cloud-init (24.1.3-0ubuntu1~22.04.1) jammy; urgency=medium * Upstream snapshot based on 24.1.3. (LP: #2056100). List of changes from upstream can be found at https://raw.githubusercontent.com/canonical/cloud-init/24.1.3/ChangeLog cloud-init (24.1.2-0ubuntu1~22.04.1) jammy; urgency=medium * refresh patches: - d/p/retain-ec2-default-net-update-events.patch * Upstream snapshot based on 24.1.2. (LP: #2056100). List of changes from upstream can be found at https://raw.githubusercontent.com/canonical/cloud-init/24.1.2/ChangeLog cloud-init (24.1.1-0ubuntu1~22.04.1) jammy; urgency=medium * d/apport-general-hook.py: Move apport hook to main branch * d/cloud-init.maintscript: remove /etc/cloud/clean.d/README * d/cloud-init.logrotate: add logrotate config for cloud-init * d/cloud-init.templates: enable WSL datasource by default * Drop d/p/retain-netplan-world-readable.patch: - Limit perms to 600 of /etc/netplan/50-cloud-init.yaml instead of 644 (LP: #2053157) * d/p/keep-dhclient-as-priority-client.patch: - keep dhclient as default client * d/p/revert-551f560d-cloud-config-after-snap-seeding.patch - Retain systemd ordering cloud-config.service After=snapd.seeded.service * d/p/retain-ec2-default-net-update-events.patch: Reverts 4dbb08f5f0cc4f41cf9dd1474f0600a11510a3c9 to not change behavior on stable releases. * d/po/templates.pot: update for wsl * d/cloud-init.postinst: change priority of hotplug rules. Avoids LP #1946003 on upgraded systems. References: [0] canonical/cloud-init#4799 [1] commit/b519d861aff8b44a0610c176cb34adcbe28df144 * refresh patches: - d/p/status-do-not-remove-duplicated-data.patch - d/p/status-retain-recoverable-error-exit-code.patch * Upstream snapshot based on 24.1.1. (LP: #2056100). List of changes from upstream can be found at https://raw.githubusercontent.com/canonical/cloud-init/24.1.1/ChangeLog

xdxu · 2024-05-01T16:40:46Z

Thanks for the reply @aciba90
I created an issue #5249 for further discussion.

aciba90 force-pushed the aws-auto-multinic branch 2 times, most recently from 0468bb1 to 0b3ae1d Compare January 29, 2024 10:24

aciba90 force-pushed the aws-auto-multinic branch from 0b3ae1d to 027848b Compare January 29, 2024 20:41

aciba90 marked this pull request as ready for review January 29, 2024 20:47

aciba90 requested review from TheRealFalcon, blackboxsw and holmanb January 30, 2024 10:41

aciba90 force-pushed the aws-auto-multinic branch from 027848b to 00b4b41 Compare January 30, 2024 12:05

TheRealFalcon self-assigned this Jan 30, 2024

TheRealFalcon requested changes Jan 30, 2024

View reviewed changes

nmeyerhans reviewed Jan 31, 2024

View reviewed changes

cloudinit/sources/DataSourceEc2.py Outdated Show resolved Hide resolved

aciba90 force-pushed the aws-auto-multinic branch from 00b4b41 to 07f8c5e Compare February 1, 2024 14:54

aciba90 force-pushed the aws-auto-multinic branch from 07f8c5e to a8215c4 Compare February 1, 2024 17:07

aciba90 force-pushed the aws-auto-multinic branch 2 times, most recently from 415c061 to 928c7b3 Compare February 5, 2024 12:37

aciba90 force-pushed the aws-auto-multinic branch from 928c7b3 to eaf420f Compare February 5, 2024 12:50

blackboxsw pushed a commit that referenced this pull request Feb 15, 2024

feat(ec2): add hotplug as a default network update event (#4799)

8110a53

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

blackboxsw pushed a commit that referenced this pull request Feb 15, 2024

feat(ec2): add hotplug as a default network update event (#4799)

085326e

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

aciba90 added a commit to aciba90/cloud-init that referenced this pull request Feb 15, 2024

d/cloud-init.postinst: change priority of hotplug rules

1d122b0

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

aciba90 added a commit to aciba90/cloud-init that referenced this pull request Feb 15, 2024

d/cloud-init.postinst: change priority of hotplug rules

9b1d4c6

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

blackboxsw pushed a commit that referenced this pull request Feb 15, 2024

d/cloud-init.postinst: change priority of hotplug rules

fefbf0b

Avoids LP: #1946003 on upgraded systems. References: [0] #4799 [1] b519d86

blackboxsw pushed a commit to aciba90/cloud-init that referenced this pull request Feb 15, 2024

d/cloud-init.postinst: change priority of hotplug rules

4ef745c

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

blackboxsw pushed a commit that referenced this pull request Feb 15, 2024

d/cloud-init.postinst: change priority of hotplug rules

22f7e45

Avoids LP: #1946003 on upgraded systems. References: [0] #4799 [1] b519d86

blackboxsw pushed a commit that referenced this pull request Feb 16, 2024

feat(ec2): add hotplug as a default network update event (#4799)

d838a55

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

blackboxsw pushed a commit to aciba90/cloud-init that referenced this pull request Feb 16, 2024

d/cloud-init.postinst: change priority of hotplug rules

154458b

Avoids LP: #1946003 on upgraded systems. References: [0] canonical#4799 [1] canonical@b519d86

blackboxsw pushed a commit that referenced this pull request Feb 16, 2024

d/cloud-init.postinst: change priority of hotplug rules

57bb323

Avoids LP: #1946003 on upgraded systems. References: [0] #4799 [1] b519d86

blackboxsw pushed a commit that referenced this pull request Feb 16, 2024

feat(ec2): add hotplug as a default network update event (#4799)

f552ddb

Now that LP: #1946003 cannot happen on ec2, because we match on NIC drivers, enable network updates on hotplug events by default on the platform.

xdxu mentioned this pull request May 1, 2024

Unexpected routing policy added in EC2 instance with multiple interfaces #5249

Open

aciba90 mentioned this pull request Jun 5, 2024

ec2, hotplug: race between udev add event and IMDS data not being available #5373

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(ec2): support multi NIC/IP setups #4799

feat(ec2): support multi NIC/IP setups #4799

aciba90 commented Jan 24, 2024 •

edited

Loading

rpocase commented Jan 30, 2024

TheRealFalcon left a comment

nmeyerhans commented Jan 31, 2024

xdxu commented Apr 29, 2024 •

edited

Loading

aciba90 commented Apr 29, 2024

xdxu commented May 1, 2024

feat(ec2): support multi NIC/IP setups #4799

feat(ec2): support multi NIC/IP setups #4799

Conversation

aciba90 commented Jan 24, 2024 • edited Loading

Additional Context

Test Steps

Checklist

Merge type

rpocase commented Jan 30, 2024

TheRealFalcon left a comment

Choose a reason for hiding this comment

nmeyerhans commented Jan 31, 2024

xdxu commented Apr 29, 2024 • edited Loading

aciba90 commented Apr 29, 2024

xdxu commented May 1, 2024

aciba90 commented Jan 24, 2024 •

edited

Loading

xdxu commented Apr 29, 2024 •

edited

Loading