- Update the module to Terraform 0.13 compatibility.
- See Upgrade Notes for TF 0.13 instructions.
- Last version compatible with Terraform 0.12
- Update Project Factory and Network modules for Terraform 0.13 upgrade.
- See Upgrade Notes for TF 0.13 instructions.
- Set instance group size to 0 when num_instances=0 (#43)
- Add
startup_script
input variable to specify a bash startup-script to execute after routing is initialized. (#40)
- (#7) Monitor the health of IP forwarding. See PR 37 for the conditions when auto-healing is triggered in response to kernel routing table changes.
- (#28) Fix Error: Invalid prefix for given prefix length
- (#32) Modify update policy to match GKE defaults. (maxSurge=1 maxUnavailable=0)
- (#35) Remove nic0_cidrs, NIC0_CIDRS, nic1_cidrs, NIC1_CIDRS from 50_compute, they are not used inside the instance by policy based routing.
- Fix issue/27
target_size
should not be set with an auto scaler.
- Pin the os image to a specific version to ensure consistent behavior when scaling in, scaling out, auto-healing, and across multiple terraform apply runs.
- Replaced the
os_image
input var withimage_project
,image_family
, andimage_name
.
- Add support and documentation for zero-downtime upgrades. See issue/23.
- Follow the process described in UPGRADE.md when upgrading to this major version.
- Set instance group size to 0 when num_instances=0 (#43)
- Add
startup_script
input variable to specify a bash startup-script to execute after routing is initialized. (#40)
- Fix issue/27
target_size
should not be set with an auto scaler.
- Fix issue/20
modules/52_regional_multinic
now deploys instance groups to all available zones in the specified region. Fixes error deploying to us-east1 and europe-west1 where there is noa
zone. - Note, resources will be destroyed and re-created. The inputs to
52_regional_multinic
have not changed relative to v1.4.0. Thezone
input to50_compute
is replaced byzones
.
- Multiple region support. See [examples/multiregion/][multiregion].
- Add
autoscale
input var, defaulttrue
, to enable auto scaling based on CPU utilization. - Changed default instance type to
n1-highcpu-2
to gain 10Gbps send rate. - Added analysis doc of autoscaler behavior at AUTOSCALER.md.
- Remove the use if iptables fwmark matches.
- Use
ip route replace
instead ofip route add
to prevent routes from piling up. - Specify ip rule priorities for clarity.
systemctl stop policy-routing
removes rules, effectively turning off policy routing.
- Workaround
systemctl restart google-guest-agent
breaking policy based routing. Resolves #10. See also guest-agent issue #76. - Send info(), debug() and error() logs to Stackdriver. Use filter
logName="projects/[PROJECT_ID]/logs/multinic"
to find them. - Add num_instances_b input to control how many instances in each zone.
- Add preemptible input var, defaults to false
- Switch routing mode from ECMP to ILB as Next Hop.
- Update to centos8.
- Virtual wire behavior, traffic ingress to eth0 egresses eth1 and vice-versa.
- Addition of hc-traffic.service unit file for LB health checks.
- Addition of hc-health.service unit file for MIG auto-healing checks.
- Added basic Cloud Logging integration example.
- Document ECMP behavior
- Document recovery time results
- Create highcpu endpoint instances
- Add troubleshooting document
- Merge branch 'improve_recovery_time'
- Use a dhcp exit hook to restore rt1
- Enable running script directly for rapid R&D
- Delete stale route resources
- Initial release