yahoo-eng-team team mailing list archive

Thread
Date

[Bug 2077455] [NEW] leftover devices causes issues while scanning new devices

To: yahoo-eng-team@xxxxxxxxxxxxxxxxxxx
From: Rajat Dhasmana <2077455@xxxxxxxxxxxxxxxxxx>
Date: Tue, 20 Aug 2024 18:08:51 -0000
Reply-to: Bug 2077455 <2077455@xxxxxxxxxxxxxxxxxx>
Sender: noreply@xxxxxxxxxxxxx

Public bug reported:

When attaching/detaching volumes, due to operation failures because of infrastructure problems like network outage, we are left with a situation where Cinder unmaps the volume from the compute host but we there are still leftover devices on the compute host.
These device cause issues for further attachment/detachment and even during scanning.
If there are leftover devices with target numbers (h:c:T:l) that are not relevant to the current connections, we keep using the same targets for scanning new devices and fail with device not found.
These entries are left in the /sys/class/fc_transport path from where os-brick fetches the channel and target values for scanning.

It would be ideal to have efforts minimizing this situation that when a
volume/LUN is unmapped to the compute node, we should have a mechanism
that check these conditions and cleans up the leftover devices.

** Affects: nova
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/2077455

Title:
  leftover devices causes issues while scanning new devices

Status in OpenStack Compute (nova):
  New

Bug description:
  When attaching/detaching volumes, due to operation failures because of infrastructure problems like network outage, we are left with a situation where Cinder unmaps the volume from the compute host but we there are still leftover devices on the compute host.
  These device cause issues for further attachment/detachment and even during scanning.
  If there are leftover devices with target numbers (h:c:T:l) that are not relevant to the current connections, we keep using the same targets for scanning new devices and fail with device not found.
  These entries are left in the /sys/class/fc_transport path from where os-brick fetches the channel and target values for scanning.

  It would be ideal to have efforts minimizing this situation that when
  a volume/LUN is unmapped to the compute node, we should have a
  mechanism that check these conditions and cleans up the leftover
  devices.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/2077455/+subscriptions