[Bug 1983499] Re: [Rbd] rados_connect_timeout has no effect

[Edward Hope-Morley]

** Also affects: cloud-archive
   Importance: Undecided
       Status: New

** Also affects: cloud-archive/yoga
   Importance: Undecided
       Status: New

-- 
You received this bug notification because you are a member of Ubuntu
OpenStack, which is subscribed to Ubuntu Cloud Archive.
https://bugs.launchpad.net/bugs/1983499

Title:
  [Rbd] rados_connect_timeout has no effect

Status in Ubuntu Cloud Archive:
  New
Status in Ubuntu Cloud Archive yoga series:
  New
Status in glance_store:
  Fix Released
Status in python-glance-store source package in Jammy:
  New

Bug description:
  [ Impact ]

  When the bug is present then the timeout has no effect, and hence it
  cannot fail back to any other backend if configured.

  The bug takes into consideration of the configuration option, and
  passes the correct values to ceph, so that the timeout is actually
  taken care of.

  [ Test Plan ]

  * Create a cloud with jammy yoga
  * Create 1 ceph backends and that as available as a backend
  * Create a ceph-rgw unit, and allow this to be a swift backend
  * Create images that point to both the rbd backend as well as the swift backend
  * Disable the default backend, i.e. cinder
  * Now creating the instance should fail, and would time out
  * Now update the package to the new release, and the timeout value to 300 for rados_connect_timeout
  * Now creating the instance should succeed

  [ Where problems could occur ]

  The previous option of the variable was not making a difference at
  all. Now that the timeout will change, and have an effect, could be
  that if the operation doesn't complete in that period is generally
  slow. it could still timeout with the new value, and use the next
  backend.

  The change only effects the timeout of ceph rbd backends, so doesn't
  cater for any other backends. So, we could still have an issue with
  timeouts on those

  [ Other Info ]

  
  We currently pass a 'timeout' argument to rados.Rados.connect (see https://opendev.org/openstack/glance_store/commit/6851cab51a3d5dbf018f755efee8e0e640ac12ab). Unfortunately, its value is silently ignored (see https://docs.ceph.com/en/latest/rados/api/python/#rados.Rados.connect).

  Users setting a value for rados_connect_timeout (which is what we
  currently pass to rados.Rados.connect) are therefore not doing
  anything.

  We should deprecate the option and encourage users to use
  client_mount_timeout in Ceph's config (but we might want a Ceph expert
  to confirm that :p).

To manage notifications about this bug go to:
https://bugs.launchpad.net/cloud-archive/+bug/1983499/+subscriptions