Handle Existing MountPoints in NodePublish and NodeStage

[davidz627]

Currently if there is an existing mount point in the place where we are trying to mount the code will just throw an error. We should confirm whether it is compatible or not and return success (for idempotency) or throw and error respectively.

[davidz627]

No longer just throws an error but does not check for compatibility

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten

[davidz627]

/remove-lifecycle rotten
/lifecycle frozen

[hantaowang]

/assign @hantaowang

[hantaowang]

The kubernetes mount interface and mount-linux has some functionality to do TODO #1 and #3 via checking /proc/mounts and proc/self/mountinfo. The difference between the two files is that

1. /proc/mounts has a mapping from device to mount target
2. /proc/self/mountinfo has a mapping from source to target and additional fields for the underlying device

I can probably get all the information I need from mountinfo, but the currently utilities that parse it are not exported and do work exactly as I want it to - so I may have to reimplement some of the logo with modifications

If we were only considering mounts from attached block devices, I can do it all with proc/mounts because the utility for praising that is already implemented well with Interface.List(). However this fails for fs mounts since proc/mounts only has a mapping from devices. So emptydir (or dirs on the host machines) will all mount from sda which does not uniquely identify the source.

[msau42]

This is specific to PD driver which can only be a block device

[hantaowang]

What is then the difference between GetBlock() and GetMount() in NodePublish? Which one is the one normally used in the case of staging and then publishing a GCE PD and when would the other be used? I see that only GetBlock() results in deriving the device path.

[hantaowang]

/cc @jingxu97