Bug 1721475

Description of problem:
======================
Raising the bug to improve the Gluster Deployment time taken in RHHI. We can remove few checks which might not be actually required.

Version-Release number of selected component (if applicable):
============================================================

How reproducible:
================


Steps to Reproduce:
==================
1.
2.
3.

Actual results:
==============

Expected results:
================


Additional info:
================

--- Additional comment from RHEL Product and Program Management on 2019-05-22 09:25:42 UTC ---

This bug is automatically being proposed for the next minor release of Red Hat Gluster Storage by setting the release flag 'rhgs‑3.5.0' to '?'. 

If this bug should be proposed for a different release, please manually change the proposed release flag.

--- Additional comment from Sahina Bose on 2019-05-22 09:36:53 UTC ---

(In reply to bipin from comment #0)
> Description of problem:
> ======================
> Raising the bug to improve the Gluster Deployment time taken in RHHI. We can
> remove few checks which might not be actually required.
>

Can you be specific on the checks to be removed?

--- Additional comment from bipin on 2019-05-22 09:45:00 UTC ---

Things noted in the current ansible based Gluster deployment:
================================================================

1.Remove the task to start firewalld
  a.TASK [gluster.features/roles/gluster_hci : Enable and start glusterd and chronyd] ***

2.Port Specific
  a.Remove the task that opens port 5666 - used by Nagio in the list of ports. These are no longer required to be opened
  b.There are ports opened from 5900-6923. Why is that required ?

3.Skipped tasks can be removed altogether ?
  a.Tasks involves NFS ganesha, CTDB, Samba are skipped, but still it takes some time to skip those. Can we totally eliminate it ?

4.Multipath configuration changes
  a.QE to test and get back with proper values

5.Glusterfs slice
  a.Ansible roles is also configures glusterfs slice
  b.Adding the host to engine, also configures glusterfs slice

6.Remove the package check step altogether


Doc link:
========
https://docs.google.com/document/d/1p0IsFo-acc7UdSXtpnUoqlVzudLwcrAJH3ATtvw1fJs/edit?ts=5ce51097

--- Additional comment from Sachidananda Urs on 2019-05-27 06:49:52 UTC ---

(In reply to bipin from comment #3)
> Things noted in the current ansible based Gluster deployment:
> ================================================================
> 
> 1.Remove the task to start firewalld
>   a.TASK [gluster.features/roles/gluster_hci : Enable and start glusterd and
> chronyd] ***
> 

Ack. Done.


> 2.Port Specific
>   a.Remove the task that opens port 5666 - used by Nagio in the list of
> ports. These are no longer required to be opened
>   b.There are ports opened from 5900-6923. Why is that required ?

These were done as part of pre-requisites, following earlier documentation.
These can be skipped by cockpit, since these are just variables. We can
configure based on which ports we need.

> 
> 3.Skipped tasks can be removed altogether ?
>   a.Tasks involves NFS ganesha, CTDB, Samba are skipped, but still it takes
> some time to skip those. Can we totally eliminate it ?

This is by design. The gluster.features role defines all the features like (rhhi, nfs, ctdb) ... 
Skipping these tasks takes less than a second.

> 
> 4.Multipath configuration changes
>   a.QE to test and get back with proper values

Ack.

> 
> 5.Glusterfs slice
>   a.Ansible roles is also configures glusterfs slice
>   b.Adding the host to engine, also configures glusterfs slice



> 
> 6.Remove the package check step altogether

Ack. Done.

> 
> 
> Doc link:
> ========
> https://docs.google.com/document/d/1p0IsFo-
> acc7UdSXtpnUoqlVzudLwcrAJH3ATtvw1fJs/edit?ts=5ce51097

--- Additional comment from SATHEESARAN on 2019-05-27 09:41:04 UTC ---

Let's try to have precise changes listed here:

1. Port to be removed from the list of ports opened: 

5666
Reason: Nagios no longer used

5900-6923. 
Why is that required ? who's using these ports ?

2. Multipath config changes.
QE to provide input based on testing

3. Glusterfs slice configuration
Decision: to be removed
Reason: Redundant.This is already taken care while adding the host in to the cluster

4. Remove package check
Decision to be remove the package check like ovirt-host, screen, vdsm, etc
Reason: RHVH bundles all these packages, and for RHEL installing these packages are 
required to be done as pre-requisite.

--- Additional comment from RHEL Product and Program Management on 2019-05-27 09:59:17 UTC ---

This BZ is being approved for the RHGS 3.5.0 release, upon receipt of the 3 ACKs (PM,Devel,QA) for the release flag 'rhgs-3.5.0', and on being attached to an official RHGS 3.5.0 BZ Tracker

--- Additional comment from SATHEESARAN on 2019-05-30 09:38:52 UTC ---


(In reply to SATHEESARAN from comment #5)
> Let's try to have precise changes listed here:
> 
> 1. Port to be removed from the list of ports opened: 
> 
> 5666
> Reason: Nagios no longer used
> 
> 5900-6923. 
> Why is that required ? who's using these ports ?
> 
> 2. Multipath config changes.
> QE to provide input based on testing
> 
> 3. Glusterfs slice configuration
> Decision: to be removed
> Reason: Redundant.This is already taken care while adding the host in to the
> cluster
> 
> 4. Remove package check
> Decision to be remove the package check like ovirt-host, screen, vdsm, etc
> Reason: RHVH bundles all these packages, and for RHEL installing these
> packages are 
> required to be done as pre-requisite.


Sac,

As part of this bug following can be treated as final requirements:
1. Remove package check
2. Remove glusterfs slice configuration
3. Remove port 5666 from the list of ports opened

For multipath configs, I will raise a separate bug

Sac, Are you  clear with the above list of requirements ?

--- Additional comment from Sachidananda Urs on 2019-05-30 09:55:54 UTC ---

(In reply to SATHEESARAN from comment #7)
> (In reply to SATHEESARAN from comment #5)
> > Let's try to have precise changes listed here:
> > 
> > 1. Port to be removed from the list of ports opened: 
> > 
> > 5666
> > Reason: Nagios no longer used
> > 
> > 5900-6923. 
> > Why is that required ? who's using these ports ?
> > 
> > 2. Multipath config changes.
> > QE to provide input based on testing
> > 
> > 3. Glusterfs slice configuration
> > Decision: to be removed
> > Reason: Redundant.This is already taken care while adding the host in to the
> > cluster
> > 
> > 4. Remove package check
> > Decision to be remove the package check like ovirt-host, screen, vdsm, etc
> > Reason: RHVH bundles all these packages, and for RHEL installing these
> > packages are 
> > required to be done as pre-requisite.
> 
> 
> Sac,
> 
> As part of this bug following can be treated as final requirements:
> 1. Remove package check
> 2. Remove glusterfs slice configuration
> 3. Remove port 5666 from the list of ports opened
> 
> For multipath configs, I will raise a separate bug
> 
> Sac, Are you  clear with the above list of requirements ?

Ack.

--- Additional comment from Sachidananda Urs on 2019-06-04 09:28:00 UTC ---

https://github.com/gluster/gluster-ansible/pull/70

Awaiting updates on multipath.

--- Additional comment from Sachidananda Urs on 2019-06-11 05:10:30 UTC ---

All the above checks will be made before any changes are done.

Copying comments from bug raised by Sean:

Original bug: https://bugzilla.redhat.com/show_bug.cgi?id=1713929

[RFE] Deployment system for RHHI-V 1.6 lacks the ability to run the prerequisite playbook in the cockpit as preflight check,so potential points of failure are made visible before deploying (with maybe link to KB article to resolve )

--- Additional comment from Sachidananda Urs on 2019-06-11 05:11:33 UTC ---

sas, I'm waiting on the multipath updates. Will move the bug to POST once those details are provided.

--- Additional comment from Sachidananda Urs on 2019-06-11 05:13:39 UTC ---



--- Additional comment from SATHEESARAN on 2019-06-14 03:49:19 UTC ---

(In reply to Sachidananda Urs from comment #11)
> sas, I'm waiting on the multipath updates. Will move the bug to POST once
> those details are provided.

I have done couple of testing paths:
1. With multipath entries for local disks in place, rebooted the host,
 to make sure that the bricks are mounted with XFS UUID. This works good.

2. Tried to deploy with multipath entries instead of disk names. This is little
tough to test with, as the deployment procedure itself flushes away the mpath entries.

It requires full fledged testing and I would recommend to go ahead with, removing
the places where multipath is configured.
Following items are done and those should be removed:
        1. Generate /etc/multipath using the command 'vdsm-tool configure --force'
        2. Edit multipath.conf to have blacklisting configuration
I have created a separate bug 1720334 to take care of this issue

--- Additional comment from Atin Mukherjee on 2019-06-16 14:40:54 UTC ---

Need an estimate on how soon this fix can land into upstream. This is to assess to current state to meet to the development freeze milestone.

--- Additional comment from Sachidananda Urs on 2019-06-18 11:23:28 UTC ---

(In reply to Atin Mukherjee from comment #14)
> Need an estimate on how soon this fix can land into upstream. This is to
> assess to current state to meet to the development freeze milestone.


PR: https://github.com/gluster/gluster-ansible-features/pull/31 fixes the issue.

This was done quite sometime ago, I forgot to update the bug. Sorry about that.