mirror of
https://github.com/hanzoai/volume.git
synced 2026-10-01 00:48:00 +00:00
A fork of Rancher's local-path-provisioner at v0.0.36. The behaviour is unchanged and deliberately so — a directory on the node a pod lands on is the right thing for a disk. What changes is identity. The provisioner answers to hanzo.ai/volume rather than rancher.io/local-path. That is not branding: two provisioners sharing one name race for the same claims, and a stock k3s carries the upstream one, so a cluster running both would settle our storage by chance. Data lives at /var/lib/hanzo/volume, so reading a Hanzo cluster does not mean learning another vendor's directory. Apache-2.0 as published, unaltered, with Rancher's attribution in NOTICE. The README says what this is not: durability. A volume here dies with its node. That is hanzoai/replicate's question, one layer up, and keeping the two apart is what lets the disk stay this simple.
28 lines
979 B
Go
28 lines
979 B
Go
// Copyright 2019 The Prometheus Authors
|
|
// Licensed under the Apache License, Version 2.0 (the "License");
|
|
// you may not use this file except in compliance with the License.
|
|
// You may obtain a copy of the License at
|
|
//
|
|
// http://www.apache.org/licenses/LICENSE-2.0
|
|
//
|
|
// Unless required by applicable law or agreed to in writing, software
|
|
// distributed under the License is distributed on an "AS IS" BASIS,
|
|
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
// See the License for the specific language governing permissions and
|
|
// limitations under the License.
|
|
|
|
//go:build (linux && appengine) || (!linux && !darwin)
|
|
// +build linux,appengine !linux,!darwin
|
|
|
|
package util
|
|
|
|
import (
|
|
"fmt"
|
|
)
|
|
|
|
// SysReadFile is here implemented as a noop for builds that do not support
|
|
// the read syscall. For example Windows, or Linux on Google App Engine.
|
|
func SysReadFile(file string) (string, error) {
|
|
return "", fmt.Errorf("not supported on this platform")
|
|
}
|