This solutions allows to generate reports on the traffic generated by subnets or hosts inside the VPC network.
Generated reports allow finding top sources of the traffic in a specific VPC.
For monitoring reasons it may be necessary to determine which hosts or networks in the VPC generate most of the traffic towards other internal hosts or the Internet. This solution allows to generate a report with desired granularity, label Google API IPs, and custom IP ranges and ports.
- Logs sink and filter (for collecting only Egress traffic)
- BigQuery dataset (for storing traffic logs)
- BigQuery view (report)
- BigQuery functions (aggregation and labelling of the addresses/ports for the view)
The following items should be provisioned before installing these reports:
- An existing project where the log sink will be created.
- An existing project where BigQuery dataset will be created.
- VPC flow logs must be already enabled in the target subnets where traffic should be monitored.
Please rename terraform.tfvars.example
to terraform.tfvars
and change the values to point to the correct log sink project ID and VPC project ID.
This project comes with a pre-populated list of IP addresses for Google APIs. These IPs change from time to time, so there is a script update-ip-range-labels.sh
which takes care of updating the list in ip-range-labels.yaml
. You need to run it manually to update the list, it is not a part of Terraform run.
This solution allows adding custom labels to specific IP ranges and ports. Edit the corresponding sections of the labels.yaml
to add the mapping between the hosts or subnets and text labels.
To label the traffic on the specific port add it under the port_labels
key.
Re-run the update-ip-range-labels.sh
script after you changed the labels.yaml
file to regenerate IP ranges definitions. The labels.yaml
file will not be read by terraform.
Instead, the generated file ip-range-labels.yaml
will be ingested by Terraform.
There are several Terraform input variables which change the report output. They do not affect thevolume of the logs exported to Big Query or the tables scanned to generate the report.
enable_split_by_destination
- set tofalse
, if you are interested only in having source IPs in the reportenable_split_by_protocol
- set tofalse
, if you are not interested in split by the protocolenable_ipv4_traffic
- if set tofalse
will exclude all IPv4 traffic from the report.ipv4_ranges_to_include
ipv4_ranges_to_exclude
- list of IPs or subnets to include or exclude. Specify single IP in8.8.8.8/32
form.ipv4_aggregate_prefix
- if the subnet is not mentioned in thelabels.yaml
, then at which granularity level aggregate traffic together. I.e. when it is24
, all IPs in10.239.1.0/24
network will be labellednetaddr4-10.239.1.0
. If you want to see per-hosts statistics, please use32
as a value.
enable_ipv6_traffic
- same as above, but forIPv6
trafficipv6_ranges_to_include
ipv6_ranges_to_exclude
ipv6_aggregate_prefix
- if the subnet is not mentioned in thelabels.yaml
the resulting label will benetaddr6-IPV6PREFIX
Please note that the current_month_*
and last_month_*
reports will process only the tables from the corresponding time ranges. If you need historical time ranges - please change the view implementation.
Once installed with the right configuration values, you'll see several views with the names top_talkers_report_{current|previous}_month_{daily|weekly|monthly}
under the newly created dataset. This dataset will automatically get populated by Cloud Operations with the VPC flow logs that are enabled in the project where the log sink resides. It may take some minutes for the first entries to appear in the dataset.
Example of the generated output:
If you enable VPC flow logs, they will be sent by default to the _Default
log sink. You can either disable the _Default
log sink (not recommended) or create an exclusion rule that skips VPC flow logs.