Skip to main content
Integration11 min read

1007 - Capstone: moving an existing site onto Plexara

A crawl copies a site you already run into a release in dev: every page at the address it answered at, every redirect as a rule, the stylesheets, scripts, images and fonts each page loads, and its forms replaced by Plexara forms. A larger site moves from its source, with its images in a media folder. The series assembled end to end: copy, review, share, promote, put the domain on it and make it public, and release it with the application behind it.

What you will take away from this lesson

ACME’s supplier information lives on a second website, suppliers.example.com, built years ago on a content management system one contractor understood. It has forty-odd pages, a contact form that emails a shared inbox, and links from every purchase order ACME has ever sent. Nobody wants to touch it, and nobody can afford to break its addresses.

This capstone moves it onto Plexara in one request and then runs it through everything the series has covered: review in dev, a supplier’s sign-off on stage, approval, the domain and the decision that makes it public, and a release tied to the application behind its form.

Learning Objectives

  1. 01Copy an existing site into a release and read what the copy reports.
  2. 02List what comes across in a crawl and what stays behind.
  3. 03Say when a site moves better from its source, with its images in a media folder.
  4. 04Replace the old site’s forms with Plexara forms during the copy.
  5. 05Run the copied site through review, sharing, approval and its domain.
  6. 06Make a site environment part of an application so they release together.
  7. 07Name every object the finished site is made of and who decided each.

Where this lesson sits

Six lessons built acme-stores from nothing. Most organizations do not start from nothing: they have websites already, on systems they would rather not renew. This capstone moves one of those and uses every lesson before it.

1000 Series: Sites

Open index

The 1000 series builds on the forms and endpoints of the 700 series and the sharing model of the 800 series: a site is where the pages people open live, with the same people, personas and approvals deciding who reaches them.

The request

Grace asked for the move in three sentences: the new site’s name, the address of the old one, and what to do about its form.

What you type

Create a site called acme-suppliers and copy https://suppliers.example.com/ into it. Replace the form on the /apply page with our supplier-onboarding form. Tell me how many pages and redirects came across and list anything you could not copy.
In the portal the same thing is Copy a site, on the site’s page. A copy started there runs in the background, shows its progress, carries on when the browser is closed, and can be stopped.

The copy

The agent created the site and copied the old one into it. A copy is a publish whose files come from a crawl, so it lands where every publish lands: in dev, as a release, private.

Copying the old site into release 1

What the agent calls. This is the exchange the agent has with Plexara on your behalf, shown for the technical reader. You ask in plain language; you never type any of it.

→ manage_site
  { "command": "crawl", "site": "acme-suppliers",
    "url": "https://suppliers.example.com/",
    "form_map": { "/apply": "supplier-onboarding" } }

← { "published": {
      "release": { "number": 1, "source": "crawl https://suppliers.example.com/" },
      "environment": "dev",
      "address": "https://acme-suppliers--dev.pages.example.com/" },
    "pages": 46,
    "redirects": 7,
    "skipped": [
      { "address": "/downloads/terms-2019.pdf", "reason": "answered 404" },
      { "address": "/partners/legacy-portal",   "reason": "redirected somewhere a release cannot point" }
    ] }

One call, and the old site is a release in dev: private, numbered, and open only to the people who manage it. The answer says what came across and, as importantly, what did not, with the reason for each. A copy that hides its gaps would be worse than no copy. The record of where a release came from keeps the address and drops anything after a question mark.

What comes across

A crawl starts at the address it is given and follows the site’s own links. For every page it copies, it also copies what the page needs in order to look the way it did.

What a crawl brings across

Every page, at its address

Each page is filed at the path it answered at, so /terms/payment/ on the old site is /terms/payment/ on the new one. Links from old purchase orders, bookmarks and search results keep working.

Redirects, as rules

An address that redirected on the old site becomes a rule in the release’s redirect map. Years of renamed pages come across without anyone listing them.

The sitemap, the robots file and the icon

Fetched whether or not a page links to them.

Stylesheets and scripts

Each one a page loads is copied. One linked with a version after a question mark is stored under its plain name, and the page keeps its link.

Images, in every size

Every candidate in a responsive image set, and images a page loads lazily as the reader scrolls.

What stylesheets load

The fonts, images and imported stylesheets named by a stylesheet, by a style block in a page, or by a style on an element.

A crawl copies what the old site answers a browser with. The content management system behind it, its admin screens and its database do not come, and are not needed: the pages are now files in a release, and the agent edits them the way lesson 1006 edits any site.

What stays behind

A copy is only trustworthy when it says what it left out. Every crawl returns a skipped list, and the reasons are specific.

What a crawl leaves behind, and says so

Links with a query string

/search?q=pallets

Not followed. A search results page is not a page of the site, and following them has no end.

Addresses that did not answer

answered 404

Reported under skipped with the status. The old site’s broken links are found on the way in.

Redirects a release cannot hold

redirected somewhere a release cannot point

Reported under skipped. The agent adds the rule by hand in the redirect map when the destination is one you want.

An address with a user name or password in it

refused

The copy is refused before it starts. Credentials never travel in a crawl.

More than a copy may hold

page and size limits

A copy stops at a page limit and a size limit and says so. A very large site is moved a section at a time, or from its source.

A larger site, from its source

A crawl copies what a site serves. A site with hundreds of pages and thousands of images moves better from its source, when you have it, with its images loaded once into a folder the site mounts.

Moving a larger site from its source

A crawl

when you have only the running site

Copies what the site serves today and keeps every address. The shared layout and the records a listing was generated from arrive as finished pages.

From the source

an archive or a bucket folder

When you hold the site’s source, the agent converts it into layouts and records and publishes it from an archive in your library or from a folder in a bucket your team has been granted, so thousands of files arrive without passing through the conversation. The release is rendered as it is published.

A media folder

images, video, downloads

Files that do not change with the pages go in a folder of your library, mounted at the path the site already uses, such as /images/. They are served on every environment with that environment’s access, sit outside releases, and a replaced file shows at once. HTML and scripts in a mount are sent as downloads, so a media folder cannot become a second, unreviewed site.

Response headers

a _headers file

A release can carry the security headers each path is served with: the content policy, the permissions policy, the framing rule. A release that declares no framing rule cannot be framed by another site, and a rollback restores the headers its release declared.

Either way the result is a release in dev, reviewed and promoted like any other, and the last step is the same: request every address in the old sitemap from dev and confirm each one answers or redirects before the domain moves.

Forms

The old site’s apply page had a form that posted to the content management system, which emailed a shared inbox. Copying that form would copy a dependency on the system being retired.

Forms on a copied site

  1. 01

    The old form is replaced, not copied

    A form on a copied page would post to a system that is being retired. The form map names, for a page’s path, the Plexara form that takes its place, and the copy frames that form where the old one was.

  2. 02

    The replacement is built from the form’s name alone

    Nothing on the old page decides where the new form posts. A copied page cannot point a form at an address of its own choosing.

  3. 03

    Submissions become rows

    The supplier-onboarding form from lesson 703 already stores every submission as a row, encrypts the tax ID, and runs an automation per submission. The contact form that emailed a shared inbox is now data.

Forms are built in lesson 703. A form whose application holds the site is answered on the site’s own address, so the application page and the form belong to one release.

From copy to cutover

With the copy in dev, the rest of the move is the series in order. Each step below is a lesson you have already read.

From copy to cutover

  1. 1

    Review the copy in dev

    Open acme-suppliers--dev and click through it. Ask the agent to compare the page list with the old sitemap and to fix what the skipped list named. Each fix is a release in dev.

  2. 2

    Place stage after dev and share it with the supplier

    A stage environment between dev and prod, shared with the supplier-blue-harbor persona. The people the site is written for read it before the public does, signed in as themselves.

  3. 3

    Require approval on prod and promote

    The site was created with a prod. Require approval on it, then promote dev to stage and stage to prod; the second promotion waits for an administrator, who sees the files and redirects it carries.

  4. 4

    Put the domain on prod

    Add suppliers.example.com to prod. It is added private, so the web team checks the site at its real name, signed in. Deasil routes the domain to your deployment and serves it over HTTPS.

  5. 5

    Make the domain public

    The owner’s decision, through Approvals. The portal names the domain before it opens it, and from then on every old address answers anyone from the new site.

  6. 6

    Retire the old system

    The content management system, its hosting and its plugins have nothing left to serve. The renewal that prompted the move does not happen.

A site as part of an application

The supplier site now fronts the supplier program: the form, the table of submissions, and the automation that runs for each one. Those were grouped into an application so they release together. The site’s prod environment joins them.

Making the site’s prod environment part of an application

What the agent calls. This is the exchange the agent has with Plexara on your behalf, shown for the technical reader. You ask in plain language; you never type any of it.

→ manage_application
  { "command": "add_part", "application": "supplier-program",
    "kind": "site", "part_id": "acme-suppliers/prod" }

← { "name": "supplier-program", "current_release": 3,
    "parts": [
      { "kind": "script", "id": "acme-supplier-onboarding" },
      { "kind": "script", "id": "acme-supplier-intake" },
      { "kind": "site",   "id": "acme-suppliers/prod" } ] }

→ manage_application
  { "command": "release", "application": "supplier-program",
    "note": "New application form fields and the page that explains them" }

A site that fronts an application should not move separately from it. As a part, the environment is released with the application: a release pins prod at the release stage serves, beside the form’s and the automation’s versions, and points prod at it, one step as a promotion would be. An application rollback puts the page and the form back together. The page that explains a field and the form that asks for it can no longer disagree. Applications are the subject of the 900 series.

An application's page in the portal at release 3: its owner, schema, page address and domain, and a Parts list grouped by kind, two automations, an asset, a webhook source and a schema, each with the version the release pins and a note where a newer version is waiting.
An application's page. Every part is pinned at a version by the current release, and a site environment added as a part is pinned the same way.

The two sites, end to end

ACME now runs two public websites on Plexara. This is everything they are made of, and who decided each part.

The two sites, object by object

acme-stores

site, built by the agent

Layouts and 304 store records. dev for the web team, stage for everyone internal and one supplier, prod behind approval and public at stores.example.com.

acme-suppliers

site, copied by the agent

Forty-six pages and seven redirects from the site it replaces, every address intact, its form replaced by a Plexara form. prod public at suppliers.example.com.

Releases

written by publish, patch, overlay and crawl

Numbered and immutable. Each has a note saying what it is for and a diff against the one before.

Promotions and rollbacks

made by people

Every move of every environment, with who made it. The public sites have only ever served releases a person promoted.

Approvals

given by administrators

One for each promotion to prod and one for each domain made public, each fixed to exactly what was asked.

Shares

set by the people who manage each site

A team on dev, a supplier’s persona and everyone internal on stage. Nobody was given an account for a website; they were given an environment.

The series, counted

The series, counted

websites, built and moved by an agent
2
request to copy 46 pages with their addresses
1
changes that reached the public unpromoted
0
servers, plugins or build tools to maintain
0

What changed for ACME

Key terms

Four terms cover moving a site onto Plexara and tying it to an application.

Key Terms

Crawl
A copy of an existing site into a release in dev: pages at their addresses, redirects as rules, and the files each page loads.
Form map
For a page path, the Plexara form that replaces that page’s forms during a crawl.
Skipped
The addresses a crawl could not copy, each with its reason. Returned with every crawl.
Site part
One environment of a site made part of an application, released and rolled back with it.