Install fails on openshift / skip blobfuse init-container if not needed.

[davidkarlsen]

What happened:

k logs csi-blob-node-24flf -c install-blobfuse-proxy
+ INSTALL_BLOBFUSE_PROXY=false
+ DISABLE_UPDATEDB=true
+ SET_MAX_OPEN_FILE_NUM=true
+ SET_READ_AHEAD_SIZE=true
+ MIGRATE_K8S_REPO=false
+ READ_AHEAD_KB=15380
+ KUBELET_PATH=/var/lib/kubelet
+ [ /var/lib/kubelet != /var/lib/kubelet ]
+ HOST_CMD=nsenter --mount=/proc/1/ns/mnt
+ nsenter --mount=/proc/1/ns/mnt cat+ grep ^ID=
 /etc/os-release
+ tr -d "
+ cut -d= -f2
+ DISTRIBUTION=debian
+ nsenter --mount=/proc/1/ns/mnt uname -m
+ ARCH=x86_64
+ echo Linux distribution: debian, Arch: x86_64
Linux distribution: debian, Arch: x86_64
+ . ./blobfuse-proxy/install-proxy.sh
+ set -xe
+ [ debian != ubuntu ]
+ echo skip install blobfuse/blobfuse2 for debian....
skip install blobfuse/blobfuse2 for debian....
+ [ false = true ]
+ [ true = true ]
+ nsenter --mount=/proc/1/ns/mnt sysctl -w fs.file-max=9000000
fs.file-max = 9000000
+ updateDBConfigPath=/host/etc/updatedb.conf
+ [ true = true ]
+ [ -f /host/etc/updatedb.conf ]
+ [ true = true ]
+ echo set read ahead size to 15380KB
set read ahead size to 15380KB
+ which awk
+ AWK_PATH=/usr/bin/awk
+ cat
+ nsenter --mount=/proc/1/ns/mnt udevadm control --reload
Failed to send reload request: No such file or directory

What you expected to happen:

It should register itself successfully

How to reproduce it:

install on openshift 4.18.x

Anything else we need to know?:

install parms:

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: blob-csi-driver
  namespace: openshift-gitops
spec:
  project: default
  syncPolicy:
    automated:
      prune: true
  source:
    chart: blob-csi-driver
    repoURL: https://raw.githubusercontent.com/kubernetes-sigs/blob-csi-driver/master/charts
    targetRevision: 1.26.3
    helm:
      valuesObject:
        linux:
          distro: fedora
        controller:
          cloudConfigSecretName: cloud-credentials
          cloudConfigSecretNamespace: blob-csi-driver
        node:
          enableBlobfuseProxy: false
          enableAznfsMount: false
          cloudConfigSecretName: cloud-credentials
          cloudConfigSecretNamespace: blob-csi-driver
          blobfuseProxy:
            installBlobfuse2: false
            installBlobfuse: false
  destination:
    server: "https://kubernetes.default.svc"
    namespace: blob-csi-driver

Why is blobfuse needed at all if one intends to use the driver with nfs?
I simply edited the deamonset and removed the init-container - and it works just fine when using NFS.

Environment:

• CSI Driver version:
• Kubernetes version (use kubectl version):
• OS (e.g. from /etc/os-release):
• Kernel (e.g. uname -a):
• Install tools:
• Others:

[andyzhangx]

@davidkarlsen this initContainer is not only for blobfuse mount, you could set -- set node.blobfuseProxy.setReadAheadSize=false to disable ReadAheadSize setting since udevadm control --reload does not work in coreos

[davidkarlsen]

thanks, now it passes.
I'm a bit confused around the OS detection though, this runs on rhcos:

k logs csi-blob-node-2ksc7 -c install-blobfuse-proxy
+ INSTALL_BLOBFUSE_PROXY=false
+ DISABLE_UPDATEDB=true
+ SET_MAX_OPEN_FILE_NUM=true
+ SET_READ_AHEAD_SIZE=false
+ MIGRATE_K8S_REPO=false
+ READ_AHEAD_KB=15380
+ KUBELET_PATH=/var/lib/kubelet
+ [ /var/lib/kubelet != /var/lib/kubelet ]
+ HOST_CMD=nsenter --mount=/proc/1/ns/mnt
+ cut -d=+  -f2
+ grep ^ID=
nsenter --mount=/proc/1/ns/mnt cat /etc/os-release
+ tr -d "
+ DISTRIBUTION=debian
+ nsenter --mount=/proc/1/ns/mnt uname -m
+ ARCH=x86_64
+ echo Linux distribution: debian, Arch: x86_64
Linux distribution: debian, Arch: x86_64     <-- but it looks like it is detected as debian?
+ . ./blobfuse-proxy/install-proxy.sh
+ set -xe
+ [ debian != ubuntu ]
+ echoskip install blobfuse/blobfuse2 for debian....
 skip install blobfuse/blobfuse2 for debian....
+ [ false = true ]
+ [ true = true ]
+ nsenter --mount=/proc/1/ns/mnt sysctl -w fs.file-max=9000000
fs.file-max = 9000000
+ updateDBConfigPath=/host/etc/updatedb.conf
+ [ true = true ]
+ [ -f /host/etc/updatedb.conf ]
+ [ false = true ]

[andyzhangx]

what's the output of cat /etc/os-release on rhcos node? we may use same logic install-proxy-rhcos.sh for debian node, in the before, it would display as rhcos

[andyzhangx]

fixed by #2125