From 8f917d692e06d0573a060107200679d76a638e50 Mon Sep 17 00:00:00 2001 From: Giuseppe Baccini Date: Mon, 2 Oct 2023 11:50:28 +0200 Subject: [PATCH] Add decision about the HA model to be used with the s3gw with Longhorn Related to: https://github.com/aquarist-labs/s3gw/issues/361 Signed-off-by: Giuseppe Baccini --- docs/decisions/0015-s3gw-ha-model.md | 31 ++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) create mode 100644 docs/decisions/0015-s3gw-ha-model.md diff --git a/docs/decisions/0015-s3gw-ha-model.md b/docs/decisions/0015-s3gw-ha-model.md new file mode 100644 index 00000000..0aed6c3e --- /dev/null +++ b/docs/decisions/0015-s3gw-ha-model.md @@ -0,0 +1,31 @@ +# s3gw ha model + +## Context and Problem Statement + +We analyzed some High Availability - HA - concepts applied to s3gw when used with Longhorn. +The final aim of the research is to identify an HA model we can use with the s3gw. + +The full HA research work conducted until now has been proposed with the +[High Availability research](https://github.com/aquarist-labs/s3gw/pull/685) Pull request. +You can find there all the rationales, motivations and the details about the tests performed. + +## Considered Options + +We identified 3 HA models: + +- *Active/Active* +- *Active/Warm Standby* +- *Active/Standby* + +## Decision Outcome + +The 3 HA models have different performances and different implementation efforts. +For our use case, the *Active/Standby* model built on top of Longhorn actually makes +the most sense and brings the "best" HA characteristics relative to implementing a +more fully active/distributed solution. + +The *Active/Standby* model is expected to work with nothing but the primitives normally +available on any Kubernetes cluster. + +for a more comprehensive explanation about this choice, please refer to the +[High Availability research](https://github.com/aquarist-labs/s3gw/pull/685) Pull request.