Bug 1860923

(In reply to Kobi Hakimi from comment #0)
...
> %include /tmp/ignoredisk.ks
> %pre
> echo "# Ignore FC drives if present" > /tmp/ignoredisk.ks
> fc=`ls -l /dev/disk/by-path/pci-*-fc-* | wc -l`
> if [ $fc -gt 0 ]
> then
> fc_ignore=`ls -1 /dev/disk/by-path/pci-*-fc-* | xargs -I {} realpath {} | tr
> '\n' ','`
> echo "ignoredisk --drives=$fc_ignore" > /tmp/ignoredisk.ks

Do we have multipath in the installer? if we do we may need to ignore also 
the multipath devices (e.g. /dev/mapper).

Can we replace the fragile ignore list with simple and more robust include
list? So we specify only the disks that should be used by the installer?

> ...
> zerombr
> clearpart --all --initlabel

Is this something that we do or part of the installer?

Changing to documentation. This is out of scope of RHV, but we should put a warning to upgrade doc about being extra careful not to wipe out SD disks

Please add a warning to Upgrade and Installation Guide:

Warning: Please be aware that you might have Data-Disks attached to the host that you could destroy if not excluded carefully. Red Hat strongly recommends to detach non-OS storage during re-installation to avoid any chance of data loss during (Re-)Installation of a host.

Here's the warning I'm adding.

[WARNING]
====
If you are reinstalling a host, {org-fullname} strongly recommends that you first detach any existing non-OS storage that is attached to the host to avoid any chance of data loss during reinstallation.
====

INSTALL GUIDES

I'll add this to the following places (in all 4 install guides):

Preparing Storage [1]

Installing Red Hat Virtualization Hosts [2]

Installing Red Hat Enterprise Linux hosts [3]

UPGRADE GUIDE

Any topics that have to do with upgrading hosts. There's a merge request for the upgrade guide as part of bug 1802650, I'll update it with the new warning.

---------------------------------------------------

[1] https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4-beta/html-single/installing_red_hat_virtualization_as_a_self-hosted_engine_using_the_cockpit_web_interface/index#Preparing_Storage_for_RHV_SHE_cockpit_deploy

[2] https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4-beta/html-single/installing_red_hat_virtualization_as_a_self-hosted_engine_using_the_cockpit_web_interface/index#Installing_Red_Hat_Virtualization_Hosts_SHE_deployment_host

[3] https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4-beta/html-single/installing_red_hat_virtualization_as_a_self-hosted_engine_using_the_cockpit_web_interface/index#Installing_Red_Hat_Enterprise_Linux_Hosts_SHE_deployment_host

Does this warning only apply to 4.4?

I added the warning to the upgrade guide in the following merge request:

https://gitlab.cee.redhat.com/rhci-documentation/docs-Red_Hat_Enterprise_Virtualization/-/merge_requests/1639

Please review

Sounds good to me.

(In reply to Nir Soffer from comment #7)
> (In reply to Kobi Hakimi from comment #0)
> ...
> > %include /tmp/ignoredisk.ks
> > %pre
> > echo "# Ignore FC drives if present" > /tmp/ignoredisk.ks
> > fc=`ls -l /dev/disk/by-path/pci-*-fc-* | wc -l`
> > if [ $fc -gt 0 ]
> > then
> > fc_ignore=`ls -1 /dev/disk/by-path/pci-*-fc-* | xargs -I {} realpath {} | tr
> > '\n' ','`
> > echo "ignoredisk --drives=$fc_ignore" > /tmp/ignoredisk.ks
> 
> Do we have multipath in the installer? if we do we may need to ignore also 
> the multipath devices (e.g. /dev/mapper).

I tried other options to exclude the FC disks as you can see in details in:
https://bugzilla.redhat.com/show_bug.cgi?id=1641863#c23
https://bugzilla.redhat.com/show_bug.cgi?id=1641863#c32

> 
> Can we replace the fragile ignore list with simple and more robust include
> list? So we specify only the disks that should be used by the installer?

yep, I need to try --onlyuse to include only the desired disks.

> 
> > ...
> > zerombr
> > clearpart --all --initlabel
> 
> Is this something that we do or part of the installer?

Its only part of the ks file

Steve if there is anything we can do to make the warning REALLY visible and not to skip while reading through upgrade guide, I would appreciate it. (Not sure about current state, did not find link to docs)

(In reply to Lukas Svaty from comment #34)
> Steve if there is anything we can do to make the warning REALLY visible and
> not to skip while reading through upgrade guide, I would appreciate it. (Not
> sure about current state, did not find link to docs)

Lukas, please take a look at the previews in the merge requests to see where I added the Warning. Let me know what you think.

The warning text is:

====
When installing or reinstalling the host’s operating system, Red Hat strongly recommends that you first detach any existing non-OS storage that is attached to the host to avoid accidental initialization of these disks, and with that, potential data loss. 
====

and for the two Kickstart examples in the Install Guides:

====
This example assumes that all disks are empty and can be initialized. If you have attached disks with data, either remove them or add them to the `ignoredisks` property.
====

Install Guides:
https://gitlab.cee.redhat.com/rhci-documentation/docs-Red_Hat_Enterprise_Virtualization/-/merge_requests/1726

Upgrade Guide:
https://gitlab.cee.redhat.com/rhci-documentation/docs-Red_Hat_Enterprise_Virtualization/-/merge_requests/1639

The doc changes do not affect any procedures, so I'm adding the NoDocsQEReview keyword.

Detaching LUNs which are not needed by the installed is the only safe way
to avoid wiping shared storage. This is not RHV upgrade or Linux issue, 
customers have wiped RHV LUNs from Windows installer.

For Linux there is a another solution that may be more practical, blacklist
the FC driver kernel module during installation. Maybe we can recommend a list
of known kernel modules that should be blacklisted during installation?

This will not help if the host is booting from SAN, in this case the only
way is tell the installer which disk should be used.

(In reply to Nir Soffer from comment #43)
> Detaching LUNs which are not needed by the installed is the only safe way
> to avoid wiping shared storage. This is not RHV upgrade or Linux issue, 
> customers have wiped RHV LUNs from Windows installer.
> 
> For Linux there is a another solution that may be more practical, blacklist
> the FC driver kernel module during installation. Maybe we can recommend a
> list
> of known kernel modules that should be blacklisted during installation?
> 
> This will not help if the host is booting from SAN, in this case the only
> way is tell the installer which disk should be used.

Steps to workaround this issue:
1. find my FC kernel module by run the following command:
  [root~]# lspci -k | grep -A 3 "Fibre Channel"
  04:00.0 Fibre Channel: QLogic Corp. ISP2532-based 8Gb Fibre Channel to PCI Express HBA (rev 02)
	Subsystem: QLogic Corp. QLE2562 PCI Express to 8Gb FC Dual Channel
	Kernel driver in use: qla2xxx
	Kernel modules: qla2xxx

2. add "module_blacklist=qla2xxx" to the kernel parameters
3. make sure the host in maintenance and FC SD active in RHVM.
4. start provision of the same host with FC disks. 

Result:
 - indeed this ignore the FC disks.
 - as a result prevent the wipe of my FC disks.
 - set the swap not on FC.

Thanks to Nir S. and Yuval T.

I merged https://gitlab.cee.redhat.com/rhci-documentation/docs-Red_Hat_Enterprise_Virtualization/-/merge_requests/1726

So now the Warning will appear in the 4.4 Installation Guide. I still need to merge the Upgrade Guide changes.

Does this bug need to remain open

The warning was published in the Upgrade Guide e.g. here:
https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4/html-single/upgrade_guide/index#Upgrading_hosts_to_4-4_4-3_local_db

It was also published in the Installation Guides, e.g. here:
https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.4/html-single/installing_red_hat_virtualization_as_a_self-hosted_engine_using_the_cockpit_web_interface/index#Installing_Red_Hat_Virtualization_Hosts_SHE_deployment_host