Resetting cluster resets user defined disk layout

[Aleksandr Shaposhnikov]

The problem: cluster reset will loose user's defined disk layout on nodes instead of keeping it.

Steps to reproduce:

1. configure cluster and change partitioning from default one (resize root partition for example).
2. deploy cluster
3. reset cluster

Observed behavior: custom partitioning layout resetted to defaults.

[Mike Scherbakov (mihgen)]

Alex, is it 6.0? Python team - please verify on current master.

[Ryan Moe (rmoe)]

I reproduced this on ISO #124

[Nikolay Markov (nmarkov)]

I would say this is a proper behavior by design, as we're resetting environment to the default state where you can configure it again. There is also one bug close to this one: https://bugs.launchpad.net/fuel/+bug/1273006

[Mike Scherbakov (mihgen)]

Nick, the whole point of Reset feature was to PRESERVE the configuration user has already done in the environment in order not to repeat the long procedure of putting all the checkboxes, roles, networks, NICs configuration AND volume-related settings. Let's get back to what is covered and what is not by Reset feature, and see how we can complete it.

[Dmitry Pyzhov (dpyzhov)]

UI will have warning in 6.1: https://bugs.launchpad.net/fuel/+bug/1446267

[Vitaly Kramskikh (vkramskikh)]

Please don't forget to revert https://review.openstack.org/#/c/177307/ after a proper fix

[Vitaly Kramskikh (vkramskikh)]

BTW, interface configuration is preserved (at least on fake UI). So fixing disks set should be difficult

[Vitaly Kramskikh (vkramskikh)]

*should NOT be difficult

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/180591

[Nikolay Markov (nmarkov)]

The thing is, this bug is a little deeper. There was this one: https://bugs.launchpad.net/fuel/+bug/1360248 , and it was fixed roughly just by rebuilding volume configuration after reset, which was done to ensure proper configuration in case disk are returned with different IDs from udev.

But, this issue has actually nothing to do with reset feature itself, it may happen with any node in case it is accidentally rebooted/reprovisioned. So, we need to solve other task - how should we match disks with ones in nailgun between reboots? Looks like we need to update agent to send some additional info and do some matching on several parameters.

And this will whole thing is supposed to be implemented together with Volume Manager refactoring.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Dmitry Pyzhov (<email address hidden>) on branch: master
Review: https://review.openstack.org/180591
Reason: No activity for more than a month

[Mike Scherbakov (mihgen)]

Per agreement between engineering, QA, product management there is a decision to move this to 8.0. We don't seem to have enough resources to fix this in current release.
It is believed that impact on a user is not that high, so we can survive with it in the current release.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/215526

[Max Stepanov (t-max-z)]

"It is believed that impact on a user is not that high, so we can survive with it in the current release."

Impact is high for new users! This bug is hidden out of sight. I'm stepping on this hoe since 5.0 release.
It is very common for a new user to reset environment several times while learning how things work.
Then reconfiguring disks every time is annoying.
At the end of the day it gives an impression that the product is half baked.

[Randeep Jalli (jallirs)]

I am hitting this bug...... is there anyway in 9.0 to preserve my disk configuration? or at least make fuel not override what i've put in to it every time a deployment fails?

[Dmitry Pyzhov (dpyzhov)]

This is an issue in our deployment flow. Looks like we'll not fit in 9.0 HCF with this bug because it needs proper discussion and design solution.

[Dmitry Guryanov (dguryanov)]

ETA: 06.06.2016

[Andrey Maximov (maximov)]

Let's explain it in Docs and, (e.g. you reset - you lose layout), adding area-docs for this.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/215526
Committed: https://git.openstack.org/cgit/openstack/fuel-web/commit/?id=716ac3ac0af058110a7687b1c4b282fef09ab2e3
Submitter: Jenkins
Branch: master

commit 716ac3ac0af058110a7687b1c4b282fef09ab2e3
Author: Dmitry Guryanov <email address hidden>
Date: Mon Jun 6 20:09:40 2016 +0300

    Remove on_node_reset callback from volume_manager

    Receiverd calls on_node_reset on stop deployment and
    cluster reset operations. In both cases we should preserve
    user-defined disks configuration. So remove on_node_reset
    callback which just resets disks configuration.

    DocImpact
    Change-Id: I9db43d0f0fa54ac4f4d626552b0553ee7a02ca69
    Closes-Bug: #1423328

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/mitaka
Review: https://review.openstack.org/353465

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/353465
Committed: https://git.openstack.org/cgit/openstack/fuel-web/commit/?id=fdc375b59c886a0392571dfb5ef7374e76e58243
Submitter: Jenkins
Branch: stable/mitaka

commit fdc375b59c886a0392571dfb5ef7374e76e58243
Author: Dmitry Guryanov <email address hidden>
Date: Mon Jun 6 20:09:40 2016 +0300

    Remove on_node_reset callback from volume_manager

    Receiverd calls on_node_reset on stop deployment and
    cluster reset operations. In both cases we should preserve
    user-defined disks configuration. So remove on_node_reset
    callback which just resets disks configuration.

    DocImpact
    Change-Id: I9db43d0f0fa54ac4f4d626552b0553ee7a02ca69
    Closes-Bug: #1423328
    (cherry picked from commit 716ac3ac0af058110a7687b1c4b282fef09ab2e3)

[TatyanaGladysheva (tgladysheva)]

Verified on 9.1 snapshot #261.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/8.0
Review: https://review.openstack.org/370643

[Alexey Stupnikov (astupnikov)]

bug #1558173 is blocking stable/8.0 CI tests. Patch is on review, we have to wait until MU3 release.

[Alexey Stupnikov (astupnikov)]

bug #1558173 was resolved, we can now test and merge the fix.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/370643
Committed: https://git.openstack.org/cgit/openstack/fuel-web/commit/?id=adaaef1eb096b936a22e11937cc8d84b9699a89a
Submitter: Jenkins
Branch: stable/8.0

commit adaaef1eb096b936a22e11937cc8d84b9699a89a
Author: Dmitry Guryanov <email address hidden>
Date: Mon Jun 6 20:09:40 2016 +0300

    Remove on_node_reset callback from volume_manager

    Receiverd calls on_node_reset on stop deployment and
    cluster reset operations. In both cases we should preserve
    user-defined disks configuration. So remove on_node_reset
    callback which just resets disks configuration.

    DocImpact
    Change-Id: I9db43d0f0fa54ac4f4d626552b0553ee7a02ca69
    Closes-Bug: #1423328
    (cherry picked from commit 716ac3ac0af058110a7687b1c4b282fef09ab2e3)
    (cherry picked from commit fdc375b59c886a0392571dfb5ef7374e76e58243)

[TatyanaGladysheva (tgladysheva)]

Verified on MOS 8.0 + MU4 updates.