Real problems. Simulated examples.
Each example below is a question an engineer might ask, the checks the agent runs, and the answer it reaches. The resources and results are made up for illustration.
// 01 Simulated example
EC2 can't reach the internet
“Our app server in the private subnet can't reach the internet. Internal traffic is fine.”
- 01Instance is running
i-0a41 runningpass - 02Security group allows outbound
sg-07c2 egress 0.0.0.0/0pass - 03Network ACL allows the return path
acl-1f9e ephemeral openpass - 04Default route to the internet
rtb-3b80 no 0.0.0.0/0fail
Root cause
The subnet's route table has no default route, so outbound traffic has nowhere to go.
Suggested fix
Add a 0.0.0.0/0 route to the NAT gateway.
// 02 Simulated example
The load balancer returns 504
“Checkout sometimes fails with 504 Gateway Timeout. Targets look healthy.”
- 01Connection errors to targets
TargetConnectionErrorCount 0pass - 02TLS handshakes to targets
TargetTLSNegotiationErrorCount 0pass - 03Targets have spare CPU
api CPU 41%pass - 04Slow requests finish inside the idle timeout
p99 74s, idle timeout 60sfail
Root cause
The slowest requests take longer than the load balancer's 60 second idle timeout.
Suggested fix
Raise the idle timeout, or move the slow export to a background job.
// 03 Simulated example
Peered VPCs can't talk
“The billing service can't reach the database in the peered VPC.”
- 01Peering connection is active
pcx-5d21 activepass - 02Route from billing to database
10.20.0.0/16 via pcx-5d21pass - 03Route back from database to billing
10.10.0.0/16 missingfail
Root cause
Only one side has a route, so replies from the database VPC have no way back.
Suggested fix
Add the billing VPC's range to the database VPC's route table, through the peering connection.
// 04 Simulated example
A private DNS name won't resolve
“Our internal database name returns NXDOMAIN from the new reporting server.”
- 01VPC DNS support and hostnames
enableDnsSupport truepass - 02Private zone is associated with this VPC
internal.example not associatedfail
Root cause
The private hosted zone isn't associated with the reporting VPC, so its resolver can't see the record.
Suggested fix
Associate the private hosted zone with the reporting VPC.
// 05 Simulated example
S3 is denied from a private subnet
“Nightly backups to S3 fail with Access Denied from the private subnet.”
- 01S3 gateway endpoint exists
vpce-8a02 availablepass - 02Endpoint is on the subnet's route table
pl-6ea54007 presentpass - 03Endpoint policy allows the bucket
backups bucket allowedpass - 04Bucket policy allows this endpoint
aws:SourceVpce vpce-11b4 onlyfail
Root cause
The bucket policy only accepts an older endpoint, so requests through the new one are denied.
Suggested fix
Allow the new endpoint in the backup bucket's policy.
Watch the agent work through these on a network map.
Open the demo