compat with octez v11.0

v11 fixes

Fixes are required to prevent the node from rejecting all baker/endorser
requests as unauthorized:

* do not use octez's docker wrapper script. It overwrites the ACL
  configuration in a way that breaks tezos-on-gke
* add ACL configuration for local IP address

bake for several adresses in same process

bake for several addresses in same process

It is possible to bake for several addresses by passing a
space-separated list of aliases to the baker/endorser daemons.

Octez v10 support

This adds necessary ACLs for v10 support.

Performance improvements + payouts with TRD

This new release of tezos-on-gke deploys a few changes related to the recent increase of activity in the Tezos network:

• remove the sentry nodes. While they provide better isolation of the tezos baking node, they cause block propagation to be delayed. In some cases, it results in missed priorities or endorsements
• move from regional storage to zonal storage. Block validation is very I/O sensitive, and the higher performance of the local storage results in faster block processing times.

Also add support for Tezos Reward Distributor in the same cluster (optional). Read doc

With the above configuration, we observe no missed endorsements or priorities with e2-standard-4 vms.

The documentation has been updated to reflect the changes.

As always, to upgrade it is recommended to do a fresh deployment and switch over the signer to the new one.

Bugfix for terraform k8s provider v2

The newest version of terraform kubernetes provider has removed the use_config_files option, so new initializations of terraform fail. This release addresses that.

v2.0

compatible with tezos version 8.0

Introduction of a new experimental active-standby mode described here

The active-standby mode is disabled by default, but feedback is welcome!

CAUTION: it is recommended to deploy a new cluster instead of upgrading. When upgrading, delete the tezos-baking-node deployment first. The tezos-baking node is now a statefulset (with just one replica unless the active-standby mode is activated), so you may end up with two baking nodes: one deployment and one statefulset. This should not be an issue, but in combination with an insecure hot baking key, it could result in two active bakers.

v2.0-rc1

• compatible with tezos version 8.0
• introduction of a new experimental active-standby mode described here

The active-standby mode is disabled by default, but feedback is welcome!

bugfix release

Fix issue with insecure baker: when using without a remote signer, the public_keys file was incorrectly formatted, resulting in inability to bake or endorse.

v1.2

Cost saving improvements:

• rolling mode for baking node: allows for a smaller default size for the persistent regional storage. The default size is now set to 15Gi for the baking node.
• route-based cluster: eliminates NAT charges

Node operation:

• Instead of a baker, you can now deploy a public node with public rpc using the same infrastructure code. The two operations are mutually exclusive. You can select history mode for your public node with a new terraform parameter history_mode.
• support for Tezos protocol upgrade such as the upcoming delphi: launch baker and endorsers for the 006 and 007 protocols at the same time. 007 baker and endorser seamlessly kick in when the chain upgrades, so the operator needs not do anything when it happens.

Monitoring:

• now alerts when private node is missing connections to public nodes
• alerts when number of public node connections drops below 5
• more configurable alerts: send k8s alerts, node alerts and signer alerts to different slack channels

Public node operation:
Instead of a baker, you can now deploy a public node with public rpc using the same infrastructure code. The two operations are mutually exclusive. You can select history mode for your public node with a new terraform parameter history_mode.

Snapshotting:

• the default variable for snapshot_url now points to our new snapshot website, https://mainnet.xtz-shots.io/rolling .
• if you want to spin up a delphinet or dalphanet cluster, you can must change the URL. Check https://xtz-shots.io for details
• the snapshotting engine now supports HTTP redirects

Others:

• when using the remote signer, the signer previously had to be online for the chain to start syncing at first deployment. This is no longer the case, but in exchange, you need to provide the public key of the baker in the terraform config.
• terraform 0.13 is now mandatory (due to introduction of some variable sanity check)

Upgrade instructions:

Since GKE does not support migration from a VPC native to a routes-based cluster, there is no upgrade path for this release and you must spin up a new cluster then point the remote signer to the new cluster.

If you are using hot baking keys, be extra careful. Tear down the old cluster before bring up the new one.

v1.2-rc1

Cost saving improvements:

• rolling mode for baking node: allows for a smaller default size for the persistent regional storage. The default size is now set to 15Gi for the baking node.
• route-based cluster: eliminates NAT charges

Node operation:

• Instead of a baker, you can now deploy a public node with public rpc using the same infrastructure code. The two operations are mutually exclusive. You can select history mode for your public node with a new terraform parameter history_mode.
• support for Tezos protocol upgrade such as the upcoming delphi: launch baker and endorsers for the 006 and 007 protocols at the same time. 007 baker and endorser seamlessly kick in when the chain upgrades, so the operator needs not do anything when it happens.

Monitoring:

• now alerts when private node is missing connections to public nodes
• alerts when number of public node connections drops below 5
• more configurable alerts: send k8s alerts, node alerts and signer alerts to different slack channels

Public node operation:
Instead of a baker, you can now deploy a public node with public rpc using the same infrastructure code. The two operations are mutually exclusive. You can select history mode for your public node with a new terraform parameter history_mode.

Snapshotting:

• the default variable for snapshot_url now points to our new snapshot website, https://mainnet.xtz-shots.io/rolling .
• if you want to spin up a delphinet or dalphanet cluster, you can must change the URL. Check https://xtz-shots.io for details
• the snapshotting engine now supports HTTP redirects

Others:

• when using the remote signer, the signer previously had to be online for the chain to start syncing at first deployment. This is no longer the case, but in exchange, you need to provide the public key of the baker in the terraform config.
• terraform 0.13 is now mandatory (due to introduction of some variable sanity check)