[RFE] Add support to execute only out-of-band cleaning steps during automated cleaning

Bug #1634345 reported by Shivanand Tendulker on 2016-10-18
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Ironic
In Progress
Wishlist
Shivanand Tendulker

Bug Description

Ironic performs automated cleaning operation in in-band way. There are couple of automated clean steps in vendor drivers (Ex.: reset ilo, reset ilo credentials etc. in iLO drivers) that can be performed out-of-band. But there is no mechanism available to disable in-band cleaning in Ironic. The operator should have flexibility to turn off the in-band automated cleaning if he does not want to execute any in-band automated cleaning step.
This will also help operators that wish to run oob clean steps to quickly complete automated cleaning and move the node back into available pool.

Changed in ironic:
assignee: nobody → Shivanand Tendulker (stendulker)

Fix proposed to branch: master
Review: https://review.openstack.org/387921

Changed in ironic:
status: New → In Progress
summary: - [RFE] Add support to perform only out-of-band cleaning for iLO drivers
+ [RFE] Add support to execute only out-of-band cleaning steps
description: updated
tags: removed: ilo-drivers
summary: - [RFE] Add support to execute only out-of-band cleaning steps
+ [RFE] Add support to execute only out-of-band cleaning steps during
+ automated cleaning
description: updated

TYranscript of discussion with dtantsur on 28/11/2016

<dtantsur> hey stendulker
<stendulker> dtantsur: regarding discussion on oob cleaning we had on review
<stendulker> dtantsur: for automated cleaning we would get only oob cleanig steps
<stendulker> list of in-band cleaning steps comes from ramdisk
<stendulker> If we want to do for both manual and automated. It is easy to get for manual steps as input from the user
<stendulker> For automated, we cannot get it.
<dtantsur> but why would you?
<stendulker> dtantsur: Unless the clean step has a definition in the Ironic code, one will not know if its oob or inband
<dtantsur> stendulker, I'm still looking forward to understand a user story behind that. Maybe you should start with a spec, as it's an important change to me.
<stendulker> say create_configuration is a manual step and has definition in ironic so we can have a extra attribute in the clean_step to indicate its in-band
<dtantsur> stendulker, what I'd like you to keep in mind, is that a list of clean steps is always provided by a user. in case of automated cleaning it's indirectly: by setting a driver and by setting driver_info/deploy_ramdisk
<dtantsur> stendulker, while a deployer can control drivers, they can't control the ramdisk set
<stendulker> dtantsur: I will write the spec, but i'm not able to understand how it can be done for automated steps by providing extra attribute in the step as the steps itself are discovered at run-time unlike manual clean steps
<dtantsur> stendulker, my comment was wrt manual cleaning. wrt automated cleaning, I don't believe we should do it at all - see above
<stendulker> when automated cleaning is invoked, disk_erase steps is reported by the ramdisk and till then ironic could be made aware of only oob steps
<stendulker> dtantsur: I was trying to cover both manual and automated
<stendulker> dtantsur: the usecase of oob cleaning applies to both manual and automated cleaning steps
<stendulker> dtantsur: For automated, operator would need to specify that using config or driver_info field
<stendulker> dtantsur: For manual steps it could be a property of cleaning step itself which could be discovered by Ironic
<dtantsur> stendulker, hmm, let's have a spec for real. I'm already confused, and I think the discussion should be recorded somewhere anyway.
<stendulker> dtantsur: the deploy ramdisk also belongs to the deployer and if he is aware that there are no cleaning steps in it, then isnt it fair to have aoob automated cleaning?
<stendulker> dtantsur: oob
<stendulker> dtantsur: ok. Thanks. I will write a spec for the same. Thank you for your inputs.
<stendulker> dtantsur: Will capture this discussion in the bug.

Ruby Loo (rloo) on 2016-11-28
tags: added: needs-spec
Changed in ironic:
importance: Undecided → Wishlist

Change abandoned by Dmitry Tantsur (<email address hidden>) on branch: master
Review: https://review.opendev.org/387921
Reason: Abandoning because of inactivity.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers