Change description

There exists a fairly narrow race condition between CSI NodeExpandVolume and a path being populated in sysfs via the Kernel.

CSI NodeExpandVolume expects all relevant paths to be present in sysfs. This is much faster and more robust than relying on Linux utilities like multipath (which is prone to timeout on busy hosts). However, there exists an oversight (not technically a bug) in CSI NodeStageVolume, which relies on the "iscsiadm" user-space Linux utility to establish, then query for iSCSI sessions. While this operation does check for session existence, it does not check for the sysfs reflection of that new session (and path) in sysfs.

This directly leads to a race between the Kernel, and any resize operation that could occur down the line.

Project tracking

Do any added TODOs have an issue in the backlog?

Did you add unit tests? Why not?

Does this code need functional testing?

Is a code review walkthrough needed? why or why not?

Should additional test coverage be executed in addition to pre-merge?

Does this code need a note in the changelog?

Does this code require documentation changes?

Additional Information