Firstly, I have Fargate tasks in private subnets of a VPC and enable NAT Gateway to get connected with ECR for pulling the images & other on-premise servers via the internet. It works perfectly. Later I setup VPC endpoints for ECR (api & dkr), S3, Secrets, logs & remove NAT Gateway, it is working for communication with AWS Services but getting the problem for communicating with on-premise servers. So I enable NAT Gateway and then my application seems working perfectly with on-premise servers. But what I am still unclear is the communication with AWS Services (ECR, S3, Secrets and CloudWatch) happens via internet or private network with VPC endpoints? Please suggest me how to debug the communications.
Thank you for your advices in advance ~
I follow Use a private subnet with internet access & I can ssh into the tasks without VPC Endpoints & NAT gateway enabled. I cannot ssh when I try with VPC endpoints method as the communication happens via private link. I still cannot ssh with VPC endpoints method and NAT Gateway enabled.
--I think I should able to ssh as NAT Gateway is enabled now.-
The VPC endpoints you are creating are specifically "Interface Endpoints". When you create an interface endpoint, AWS
will add an elastic network interface (ENI) to your specified subnets and assign it a private IP address in your subnet's address space. In general,
you'll also tell AWS to add a DNS entry for that ENI which resolves the service's domain name against that IP (insetad
of the public IP). You can disable this, but it kind of defeats the purpose.
This effectively means that anytime you try to resolve the hostname for that service, it should resolve to your ENI's
IP address and thus go over privatelink. However, it is important to note that you need to configure your CLI/SDK for the region your ENI is in. Otherwise, it may use the generic DNS entry (which may point to us-east-1 specifically). That will resolve just fine (thanks to your NAT Gateway), but if you are running in another region, your traffic may route unexpectedly over the internet.
All of this is independent of SSH. Remember, VPC Interface Endpoints are only used to create a private IP address that
can be used to route to AWS services. If you are trying to SSH into a Fargate task, that task just needs to be routable. In this particular case, your Fargate tasks are running in your VPC, and are apparently directly routable. No NAT Gateway or interface endpoints should be necessary to reach them.
Related
I've a small Fargate cluster with a service running and found that if I disable the public IP the container won't build as it doesn't have a route to pull the image.
The ELB for ECS Fargate is part of a subnet which has:
internet gateway configured and attached
route table allowing unrestricted outgoing
security policy on the ECS service allows unrestricted outgoing
DNS enabled
My understanding is the internet gateway is a NAT and the above actions should permit outgoing internet access however I can't make it so. What else is missing?
Just like all other resources in your AWS VPC, if you don't attach a public IP address, then it needs either to be placed in a subnet with a route to a NAT Gateway to access things outside the VPC, or it needs VPC endpoints to access those resources.
I have set-up a EBL for a persistent public & subnet IP. As far as I
can tell my subnet has outgoing internet unrestricted (internet
gateway attached and route opens up all outgoing traffic to 0.0.0.0/0.
I'm unsure if the service setup will configure the EC2 to use this
first then attempt to set-up the container. If not then it probably
doesn't apply.
ELB is for inbound traffic only, it does not provide any sort of outbound networking functionality for your EC2 or Fargate instance. The ELB is not in any way involved when ECS tries to pull a container image.
Having a volatile public IP address is a bit annoying as my
understanding is the security policy will apply to both the
ELB/Elastic provided IP and this one.
What "security policy" are you referring to? I'm not aware of security policies on AWS that are applied directly to IP addresses. Assuming you mean the Security Group when you say "security policy", your understanding is incorrect. Both the EC2 or Fargate instance and the ELB should have different security groups assigned to them. The ELB would have a security group allowing all inbound traffic, if you want it to be public on the Internet. The EC2 or Fargate instance should have a security group only allowing inbound traffic from the ELB (by specifying the ELB's security group ID in the inbound rule).
I want to point out you say "EC2" in your question and never mention Fargate, but you tagged your question with Fargate twice and didn't tag it with EC2. EC2 and Fargate are separate compute services on AWS. You would either be using one or the other. It doesn't really matter in this case given the issue you are encountering, but it helps to be clear in your questions.
I'm wondering if anyone can help with my issue, here's the setup:
We have 2 separate kubernetes clusters in GKE, running on v1.17, and they each sit in a separate project
We have set up VPC peering between the two projects
On cluster 1, we have 'service1' which is exposed by an internal HTTPS load balancer, we don't want this to be public
On cluster 2, we intend on being able to access 'service1' via the internal load balancer, and it should do this over the VPC peering connection between the two projects
Here's the issue:
When I'm connected via SSH on a GKE node on cluster 2, I can successfully run a curl request to access https://service1.domain.com running on cluster 1, and get the expected response, so traffic is definitely routing from cluster 2 > cluster 1. However, when I'm running the same curl command from a POD, running on a GKE node, the same curl request times out.
I have run as much troubleshooting as I can including telnet, traceroute etc and I'm really stuck why this might be. If anyone can shed light on the difference here that would be great.
I did wonder whether pod networking is somehow forwarding traffic over the clusters public IP rather than over the VPC peering connection.
So it seems you're not using a "VPC-native" cluster and what you need is "IP masquerading".
From this document:
"A GKE cluster uses IP masquerading so that destinations outside of the cluster only receive packets from node IP addresses instead of Pod IP addresses. This is useful in environments that expect to only receive packets from node IP addresses."
You can use ip-masq-agent or k8s-custom-iptables. After this, it will work since it will be like you're making a call from node, not inside of pod.
As mentioned in one of the answers IP aliases (VPC-native) should work out of the box. If using a route based GKE cluster rather than VPC-native you would need to use custom routes.
As per this document
By default, VPC Network Peering with GKE is supported when used with
IP aliases. If you don't use IP aliases, you can export custom routes
so that GKE containers are reachable from peered networks.
This is also explained in this document
If you have GKE clusters without VPC native addressing, you might have
multiple static routes to direct traffic to VM instances that are
hosting your containers. You can export these static routes so that
the containers are reachable from peered networks.
The problem your facing seems similar to the one mentioned in this SO question, perhaps your pods are using IPs outside of the VPC range and for that reason cannot access the peered VPC?
UPDATE: In Google cloud, I tried to access the service from another cluster which had VPC native networking enabled, which I believe allows pods to use the VPC routing and possibly the internal IPs.
Problem solved :-)
Can I route requests to GKE private master from another VPC? I can’t seem to find any way to setup GCP router to achieve that:
balancers can't use master ip as a backend in any way
routers can't have next-hop-ip from another network
I can't (on my own) peer different VPC network with master private network
when I peer GKE VPC with another VPC, those routes are not propagated
Any solution here?
PS: Besides creating a standalone proxy or using third-party router...
I have multiple gcp projects, kube clusters are in separate project.
This dramatically changes the context of your question as VPC from other projects aren't routeable by simply adding project-level network rules.
For cross-project VPC peering, you need to set up a VPC Network Peering.
I want my CI (which is in different project) to be able to access private kube master.
For this, each GKE private cluster has Master Authorized Networks, which are basically IP addresses/CIDRs that are allowed to authenticate with the master endpoint for administration.
If your CI has a unified address or if the administrators have fixed IPs, you can add them to these networks so that they can authenticate to the master.
If there are not unified addresses for these clients, then depending on your specific scenario, you might need some sort of SNATing to "unify" the source of your requests to match the authorized addresses.
Additionally, you can make a private cluster without a public address. This will allow access to the master endpoint to the nodes allocated in the cluster VPC. However:
There is still an external IP address used by Google for cluster management purposes, but the IP address is not accessible to anyone.
not supported by google, workarounds exists (but they are dirty):
https://issuetracker.google.com/issues/244483997 , custom routes export/import has no effect
Google finally added custom routes export to VPC peering with master subnet. So the problem is now gone, you can access private master from different VPC or through VPN.
I use external docker registry in my GKE cluster to pull containers. This docker registry has security access list (which is basically list of client's public IP addresses). However, GKE cluster creates nodes with ephemeral IP addresses and makes it inconvenient to add each new IP address into access list.
How can I create proxy or port forwarding in google cloud to forward all requests via one external IP to access my external registry ?
You should use Cloud NAT. It will act at the proxy in your diagram and you can use the addresses of the NAT in the ACLs of the container registry. Also check out this tutorial on setting up Cloud NAT with GKE.
I have mongodb hosted in Mongo Atlas. There for security reasons I have whitelisted some IPs, that can access it.
Now that I have set up kubernetes cluster in google cloud, how can I make it able to access this mongodb service since I don't have clear IP for my cluster/swarm instances that get spawned on demand.
There is no way to get static IPs with GKE. You need to use a NAT gateway. You can configure a GCE VM to act as your NAT gateway for all outbound traffic from your cluster.
There have been multiple requests for a GCP native NAT feature and I believe that feature is on the way. In the meantime, a GCE VM acting as a NAT gateway is your best bet
EDIT you can now use Google Cloud NAT to assign a single (or multiple) static IPs to your cluster (or other Google resources)