Bug 881866

Summary: [RFE] Modify "help" package deployment.
Product: Red Hat Enterprise Virtualization Manager Reporter: Stephen Gordon <sgordon>
Component: RFEsAssignee: Scott Herold <sherold>
Status: CLOSED WONTFIX QA Contact: yeylon <yeylon>
Severity: low Docs Contact:
Priority: unspecified    
Version: unspecifiedCC: dfediuck, ecohen, gklein, iheim, jbiddle, lpeer, mkenneth, rbalakri, Rhev-m-bugs, rlandman+disabled, rlandman, sbonazzo, sgordon, srevivo, yeylon, zdover
Target Milestone: ---Keywords: FutureFeature
Target Release: ---Flags: sgrinber: Triaged+
Hardware: Unspecified   
OS: Unspecified   
Whiteboard: integration
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2014-10-28 13:07:43 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On: 870201    
Bug Blocks:    

Description Stephen Gordon 2012-11-29 17:31:19 UTC
Description of problem:

For 3.x we provide a context sensitive help mechanism. It is driven by two CSV files which link GUI elements to HTML file names and anchors. Currently the package is built by having rhevm-doc BuildRequires set on all of the documentation packages. These packages are installed on the build host and thn the spec file copies them into the rhevm-doc build root.

Other than being pretty poor form (because our source package is basically useless) it means we have to introduce a number of hacks to ensure we pick up the correct documentation, strip elements like javascript that only apply when being delivered in a publican site, etc.

On top of that we have to do the same for all localized versions which results in a lot of tagging, editing spec files, rebuilding, etc.

Recently I have filed an RFE to get publican 3 into RHEL 7 (Bug # 870201). Theoretically what this allows (and I am proposing) is for us to ship the exact (hopefully) same packages to customers as we use to display the staging server and ultimately access.redhat.com (look and feel would be like the documentation stage but we can also customize this for RHEV with a brand). This means:

- rhevm-doc to become a much smaller package, simply providing the CSV files and the publican site framework.
- rhevm-doc-en-US to be a shell that requires the "real" documentation packages.
- rhevm-doc-ja-JP to be a shell that requires the "real" Japanese documentation packages.
- and so on.

The main benefits of this are:

- More consistent look and feel of the deployed docs (3.0 was fine but 3.1 was a bit off due to hacks I had to use to get the navigation working).
- Removes the most risky hacks in rhevm-doc builds (and the other new packages being introduced wouldn't need them).
- Allows us to rely on publican's "fallback" logic when content is untranslated rather than symbolic links.
- We achieve my original goal of deploying the same packages (single sourcing). The reason we couldn't do this originally is packages built on publican > 2.5 (which we have used for quite some time) wont deploy on sites managed by publican < 2.5, RHEL 6 only shipped 2.1.

I would also like to remove the dependency on rhevm itself to allow users to install the packages on systems that are not running rhevm. Obviously all systems that *are* running rhevm would still require it to provide context sensitive help (so rhevm requires rhevm-doc, but rhevm-doc does not require rhevm). To do this I think we would also need to update the location of the docs to /usr/share/docs/ovirt-engine.

The main downsides of this are:

- The total size of an installation (rhevm-doc, rhevm-doc-en-US, and dependencies) would be larger as they would also include html-single and epub formats.
- We haven't eliminated the rel-eng overhead in the initial setup/package creation of each package.

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

I'd like to target this for RHEV 4/RHEL 7 if approved (obviously the RHEL 7 component is publican 3 inclusion, which now appears to be likely).

Comment 1 Stephen Gordon 2012-11-29 17:40:10 UTC
> To do this I think we would also need to update the location of the docs to /usr/share/docs/ovirt-engine.

NB: Just to highlight publican's default location for the "site" is /var/www/html/docs but I believe this is easily modified using a flag to the publican create site logic (which would be done in the rhevm-doc package).

Comment 2 Stephen Gordon 2012-12-18 14:24:02 UTC
Note that the suggested deployment mechanism here will need to co-operate with the changes being made under Bug # 885823.

Comment 3 Itamar Heim 2014-01-16 12:12:09 UTC
stephen - pinging back - what/who needs to do something on this one?

Comment 4 Stephen Gordon 2014-01-16 12:15:48 UTC
Hi Itamar,

I expect engineering (you, Kiril?) will need to own this but would need to work closely with ECS as they have more familiarity with how the docs packages themselves work (not necessarily rhevm-doc). 

Thanks,

Steve

Comment 5 Itamar Heim 2014-01-16 13:44:13 UTC
joey - can you please provide/liaise - what changes are planned, and what do we need to change to accommodate them?

Comment 6 Stephen Gordon 2014-01-16 14:03:51 UTC
(In reply to Itamar Heim from comment #5)
> joey - can you please provide/liaise - what changes are planned, and what do
> we need to change to accommodate them?

I think the key here is to sit down as a group (RHEVM engineering, ECS writers, and ECS Publican gurus (Rudi)) and revisit the overall process used for building rhevm-doc and see what can be done to make it more robust. Looking at the GSS plugin it seems like they are inlining these docs in the UI as well?

Key goals should be:

a) Removing the need to perform complicated sed operations in the SPEC to make the docs look OK.

b) Ensuring the SRPM is better behaved (that is, isn't just a shim for copying in files from the packages it depends on).

c) Ideally, using the same packages as used for access.redhat.com or at least packages generated from these builds with minimal additional author overhead. This was always the goal but discrepancies between the version of Publican in RHEL and the version of Publican used by ECS have made it a moving target.

Comment 7 Jodi Biddle 2014-01-17 00:05:43 UTC
Itamar - we're hoping to get over to TLV in person in a few weeks, where Zac will be able to discuss this in person and provide a better solution. I understand that Steve and Zac spoke about this at length yesterday, and we have a few options available. 

Leaving the needinfo so it doesn't fall off my radar.

Comment 9 Sandro Bonazzola 2014-08-08 11:13:11 UTC
(In reply to Jodi Biddle from comment #7)
> Itamar - we're hoping to get over to TLV in person in a few weeks, where Zac
> will be able to discuss this in person and provide a better solution. I
> understand that Steve and Zac spoke about this at length yesterday, and we
> have a few options available. 
> 
> Leaving the needinfo so it doesn't fall off my radar.

I guess we need 2 bugs, one on doc team to provide the docs (this one) and one on integration to be opened once it's ready.

No activity on this bug in last 7 months, what's the status?

Comment 11 Doron Fediuck 2014-10-28 13:07:43 UTC
Closing old RFEs.

Please re-open if still relevant while considering comment 9.