RDoc-4094 Cloud docs update for the September 2026 deployment

[PFYasu]

Documents the user-facing changes in the September deployment window.
Ticket: https://issues.ravendb.net/issue/RDoc-4094

Alerts

• New Alerts section in Maintenance & Troubleshooting: daily summary vs. custom alerts, subscribers, the per-product limits (25 rules, 20 recipients, one daily summary) and the four cases where rules are removed automatically — including a last-recipient removal deleting the whole rule.
• Trigger events point at the existing Labels table rather than restating them, since that table already carries the thresholds.
• Short generic View the Alerts section on the Products page, following the Metrics / Cluster Health pattern (generic in the Portal pages, detail outside).

Billing & Costs

• Current billing period costs donut, the Unbilled total and the Cost Details entry points.
• Spending: the cost-category filter (all 18 categories), the year-picker behaviour and the "not calculated yet" states.
• New Cost details section for /billing/cost-details, which was previously undocumented, including the prediction caveat on unbilled periods and the marketplace notices.

Portal navigation

Rewrote the Portal overview for the collapsible sidebar: the seven role-gated views, the Help center menu (Support, Documentation, Community, Service status) and the profile entry. Support and Service status are no longer top-level entries. Added the missing Users & Access entry, and noted that the account switcher identifies accounts by domain name.

Support

Request support for this instance from the product view (paid support only, pre-selects the product), plus the support resources panel on the product view.

Login

Google, GitHub and Microsoft sign-in, with each provider's email-verification requirement. Microsoft work/school (Entra ID) accounts additionally need the tenant to attest domain ownership via the xms_edov optional claim — worth calling out, because a tenant that does not emit it fails with "could not be verified" even though the address is correct. Also lists the actual sign-in failure messages.

Provisioning

Noted that the Account and Billing wizard stages are skipped when the account already has those details, and are never shown for marketplace accounts — so the documented five-stage flow is the first-product case.

---

Second pass — validated against hr/test

Every claim below was checked against the cloud portal source rather than inferred, and the whole set was re-validated after the September merge landed on hr/test.

Autoscaling — rewritten (RDBCL-3989)

RDBCL-3989 replaced the flat CPU thresholds with three independent metrics: load average, available memory and IO wait, each with its own low/high limit and duration. The old section described CPU only, and 4 of its 10 table rows no longer exist in the UI.

• Separate Upscaling and Downscaling tables, each row carrying the UI's enforced min/max from autoscalingThresholdFields.ts.
• New Autoscaling range subsection for lowest/highest type.
• Admonition on the asymmetry the code comments call out: downscaling watches the low load average and low IO wait but the high available memory, because available memory runs opposite to utilisation.
• New Serverless Autoscaling subsection — the one ProductFeatureName value that had no coverage. Always enabled (canDisableFeature={false}), and its modal exposes only the instance range; the thresholds are managed for you.

Automatic Storage Extension — corrected

• The threshold range is 70–95, not 0–100. The <input> carries min=0 max=100, but both UpdateThresholdSchema.ts (Yup.number().min(70).max(95)) and ProductValidator.cs enforce 70–95. Entering 50 was rejected with no explanation in the docs.
• Two gating rules were undocumented (ModelExtensions.cs): the product must be Active, and the threshold cannot be changed at all on yearly contracts.
• The setting is disk usage, not memory — the old wording came from a screenshot predating the Memory Used Threshold → Disk Space Used Threshold rename.

Products tab — restructured

• Security and Allowed IPs split into separate sections; they are separate cards and are not adjacent in the UI.
• Documented the Certificate expiration date card (RDBCL-4473) and the renewal schedule behind it: certificates are issued for 5 years, a renewal email goes out at 60 days, and automatic regeneration happens at 30 days (Constants.Certificate, ProvisionProductCommandHandler).
• Instance Access rewritten — ProductEditAccessSection is dead code on hr/test; Open Studio moved to the product-details top bar and becomes a per-node dropdown on multi-node products, so "opens the Studio of node A" was wrong for clusters.
• New Subscription & support section (RDBCL-5326): subscription details, the Update available release-channel pill, the marketplace guard, and the Support-role gating on the badge and CTA.
• Section order now follows the UI, and figure numbering was reconciled.

Support — response times and the skipped step

• The response-time admonition stated two conditions; the code has three. Severity is the dominant gate and was entirely undocumented: Within 2 hours additionally requires severity === "Critical", Next business day requires Normal or Critical. A paying customer filing a Minor issue finds both greyed out. Replaced with a per-option condition table.
• The tier restriction applies only to product-related questions — a general question is never tier-blocked.
• Both step lists were missing a mandatory step: the Resources that may help you panel gates the rest of the form via isFamiliarWithResources, so the Expected Response Time section does not render until you click through it. Lists are now 7 and 10 steps, using the UI's own field labels.

Backup

The mandatory backup is a Server-Wide Backup (BackupHelper takes a ServerWideBackupConfiguration), so it covers every database rather than being per-database. The navigation said "Manage Ongoing Tasks"; the view is Ongoing Tasks.

Link and alt-text hygiene

• 10 broken in-page anchors fixed across cloud/ — verified old-broken / new-valid against the rendered HTML, not guessed slugs.
• 152 image tags normalised. !["Figure N - X"](…) put the literal quotes inside the alt attribute (alt=""…""), which screen readers announce. The Figure N prefixes were invisible (LazyImage renders no caption) and referenced by no prose, while costing renumbering on every insertion. Now ![X](…). Also repaired a malformed tag in cloud-aws-marketplace.mdx that was missing its closing quote.

---

Third pass — folding in RDoc-3556

RDoc-3556 was closed as Obsolete together with its
PR #2393, which never merged. Only three of its claims (the 70–95 range, the yearly-contract block, and the
Memory → Disk correction) were already covered by the second pass above; the rest was not in the repo at all.
Its content is folded in here, re-validated against hr/test (83c55efe7) rather than carried over as written —
one claim did not survive and several gaps in it were filled.

Automatic Storage Extension — the rest of the section

• Availability. Development and Production tiers, excluding Dev10 and PN-type instances
  (ProductFeaturesHelper.IsFeatureAvailable). Also noted that an ineligible product has no feature row at all —
  GetProductFeaturesDto skips the row when the feature is absent from the PPC (RDBCL-5096), so "I can't find the
  setting" and "the setting is off" look identical in the UI.
• Enabled by default at 90%, and free. Verified both halves: RDBCL-4266 added it enabled by default and the
  20260317 db-patch backfilled existing products at 90; GetFeaturePercentage returns 0 for
  AutomaticStorageExtension, so the feature carries no charge and only the larger disk is billed. RDoc-3556 better docs for automatic storage extension #2393 asserted
  "free of charge" without this, and there is an AutomaticStorageExtensionFeature cost line type, which makes
  the claim look wrong until you follow the percentage.
• How it works. The node reports usage to /api/v2/node/storage/auto-extend; the disk then moves to the
  next size the instance type offers, not to the size that would merely clear the threshold — worth stating,
  because the jump is often much larger than the overflow. The success email carries the old, new and maximum
  size (AutoStorageExtend.html).
• Azure Premium disks also gain the new size's IOPS and throughput when the target performance tier is
  better (ExtendProductStorageVolumeHelper). Undocumented anywhere, and it changes the bill.
• When storage is not extended. RDoc-3556 better docs for automatic storage extension #2393 listed three cases from a guess at the enum; StorageExtensionForbiddenReason
  has five, and they do not map one-to-one onto emails. Documented the four that notify you — instance-type maximum,
  root-disk limit, yearly contract, Free tier — and left out StorageSizeLocked (PN instances), which sends nothing
  and cannot be reached anyway, since PN products are not eligible. Also corrected the framing: the root-disk limit
  (1,000 GB AWS/GCP, 1,024 GB Azure) applies only to root-only disk layouts, not to every product.
• One warning per day. EmailNotificationPolicy re-registers each of these notifications at most once per
  24 hours per product.

Storage modification rate — new, and not from #2393

Constants.StorageChangeLimitExpirationMinutes enforces 6 h on AWS and 12 h on Azure Premium between two
storage changes. It gates CanChangeStorage and CanExtendStorageAutomatically alike, so it silently delays an
automatic extension: a disk filling up fast can sit above its threshold for hours. Documented in the feature section
and, since it applies to manual resizes too, on Scaling → Change Storage.

Also

• Cross-link from Scaling → Change Storage to the feature (RDoc-3556 better docs for automatic storage extension #2393 added this too).
• Explicit {#storage-extension-configuration} slug on the new Configuration heading — the page already has three
  Configuration sections, so it would otherwise have become configuration-3.
• Anchors verified against the rendered HTML, as in the second pass. npm run build:current clean.

---

Before merge

• cloud/portal/assets/portal-product-alerts.png is a placeholder. Component-level capture with no portal chrome, so it does not match the framing of the neighbouring Metrics / Cluster Health figures. Needs a real-portal shot (sidebar + product header + Alerts tab active).
• Sections still without figures: Cost details, the Alerts subsections in Maintenance & Troubleshooting, and Serverless Autoscaling.
• Stale figures: portal-billing-costs-tab-current-billing-period.png still shows the 3% marketplace commission notice removed by RDBCL-4746; portal-billing-costs-tab-billing-history.png predates two column reworks (RDBCL-4739, RDBCL-4751). SNMPv3 Monitoring credentials has no figure while its sibling does.
• Terminology. The prose describes a sidebar and calls them "views", while the eight portal page titles and filenames still say "Tab". Renaming needs scripts/redirects.json entries, so it is out of scope here.

⚠️ Deploy ordering

Two features documented here are on hr/test but not on hr/prod:

Feature	hr/test	hr/prod
Autoscaling metrics (RDBCL-3989)	yes	no
"Sign in with Microsoft" (RDBCL-5283)	yes	no

Correct for the September deployment; wrong if this merges before that ships. Please check the deploy order before merging.

Noted while working

#manage-product-the-security-tab exists on no page, so three guides were landing readers at the top of the Products page. Fixed here — and not with the anchor first suggested: the link text is "For more details on adding a new product", so the target is #provisioning-a-new-product, not the security section. Each link keeps its existing absolute or relative form; only the fragment changed.

• guides/connecting-c-application-to-ravendb-cloud.mdx
• guides/connecting-node-js-application-to-ravendb-cloud.mdx
• guides/how-to-setup-ravendb-with-asp-net-core-8-application.mdx

(The previously-noted #cloud-account-portal-products from cloud/cloud-migration.mdx is fixed here too.)

Docusaurus validates page links but not anchors, which is why the build stays green either way — every anchor in this PR was checked against the rendered HTML instead.

Still unreferenced, and orphaned before this work: cloud/assets/portal-product-details-storage-upgrade-to-premium-v2.png, a duplicate of the portal/assets copy.

Verified with npm run build:current — clean, no warnings, no broken links.

---

Asset naming convention

Filenames now read <surface>-<view>-<region>[-<state>], so the name is the path you navigate to re-shoot the figure. 75 files renamed, all as git mv.

• Six pointed at the wrong place: portal-tabs (it is the sidebar, not tabs), portal-support-subscriptions-tab (it is on the product view), provider-region-tier-selection (the Plan stage, completing the five-stage wizard set), plus the two emails, now under an email- prefix.
• The 38 positionally-numbered <page>-NNN-<desc> names are retired. That scheme had already collided twice — two pricing-001 files with no pricing-002, and two tiers-and-instances-0034 — which is the same defect as the Figure N alt-text prefixes this PR already removed. Redistributed by real capture surface: portal-, studio-, os-windows-, browser-.
• -tab dropped from the seven portal groups (30 files), matching the prose that now says views. The four view overviews are consistent: portal-{account,api,home,support}-overview.

Renames need no redirect entries: scripts/redirects.json has no asset paths, built asset URLs are content-hashed, and nothing outside cloud/**/*.mdx references these files.

Duplicate and orphan assets

Six basenames existed in both cloud/assets and cloud/portal/assets, three already drifted apart. Consolidated onto the portal copy and repointed the cloud/*.mdx references — which also retires a stale-chrome copy of portal-product-list-manage-button. One unreferenced asset deleted.

Screenshots

• The cluster-health figure had the Incidents History panel captured twice; now one panel, with 2026 dates.
• The backup-task figure now shows Ongoing Tasks, matching the prose, and evidences the 14-day retention policy the page claims.
• The tier and instance figures are off the pre-redesign theme, so the product-create Plan set is uniform. Production tier limits rose to 64 cores / 256 GB RAM / 4000 GB disk, and the instance cards now carry IOPS and throughput.
• The pricing figures show list rather than discounted prices, and surface Public IP, backup and per-feature costs that Additional Expenses does not yet cover — see Follow-ups.

Figure/prose corrections found by re-reading the figures

• cloud-scaling — the IO-limits admonition said the Disk Throughput is lower than the Instance Throughput. It is the other way round in the figure, and was in the previous figure too, which inverted the point of the advice that follows it.
• cloud-overview — the Open Studio walkthrough quoted a certificate name ("free.serez") that no figure has shown for some time; the instruction now points at the URL without quoting an example.
• cloud-pricing-payment-billing — the prose compared PB10 and PB20 while the figures are PB10 and PB30.
• Alt text — the email-notification figure in cloud-scaling was labelled "Scaling storage", and two billing figures were labelled PayNow1 / PayNow2.

Follow-ups, not in this PR

Screenshot values in these figures come from the test environment and are not a concern.

Stale figure content

• portal-product-details-serverless.png is the last figure on the pre-redesign product view — Manage / Metrics / Cluster Health as buttons, no Alerts tab, and the dead "Instance access" card. It contradicts the refreshed portal-product-details.png on the same page.
• portal-product-create-customize-section.png shows release channels LTS 5.4 / Stable 6.2 / Current (7.1); CURRENT_VERSION is 7.2.
• studio-cluster-view-with-additional-nodes.png shows RavenDB 6.2.0.
• portal-billing-costs-current-billing-period.png carries 2025 dates beside a September-2026 sibling; portal-billing-costs-billing-history.png has no 2026 in its year picker.
• Both studio-notification-* figures are dated 14 July 2019; portal-product-details-audit-logs.png is December 2024; portal-api-create-api-key-modal.png shows an expiration already in the past.

Figures that show a control but not a result

• product-features-powerbi-configuration.png — IP table is empty, headers only.
• portal-product-details-cluster-health.png — Instance changes in time is a blank plot area taking roughly 30% of the image height.
• registration-page-thank-you.png — roughly 70% empty canvas.

Prose to check against the figures

• cloud-scaling.mdx:133 states Azure Standard disks are 500 IOPS and 60 MB/s, while the adjacent figure lists 500 IOPS / 100 MBps (default) in the Premium Performance tier dropdown. Same IOPS, different throughput — these may describe different things, but a reader will take them as one claim. The AWS equivalent (3000 IOPS / 125 MB/s) does match its figure.
• Three login and registration figures still show the "Captcha was disabled by configuration option" banner, a state production never renders.
• The registration walkthrough disagrees with itself across steps: the billing step gives the company as YasuIT, the summary says RavenDB.

Content the figures now show and the prose does not

The refreshed pricing figures surface Public IP $0.005/hour, backup at $1.00/GB/month after the free 1 GB, and PowerBi / OLAP ETL / Queue Sink at 10% of hosting cost — none covered by Additional Expenses. The first is what RDoc-2661 and RDoc-2662 ask for.

Separately, verify the hourly prices in portal-product-create-plan-pricing-pb10/pb30.png against current pricing. It is the only remaining item that could be actively wrong for a customer rather than merely dated.

Near-duplicate pairs the renaming exposed

Two Studio views each exist as two crops used on two different pages: studio-server-dashboard-manage-certificates{,-link}.png and studio-certificates-{client-certificate-menu,generate-client-certificate}.png. Each pair is a candidate to collapse, but that is a decision about the pages rather than a rename.

Other

• The autoscaling IO wait min value field renders empty while the summary beside it shows 0% — no RDBCL ticket yet, and product-features-autoscaling-configuration-edit.png ships the behaviour as if correct.
• 66 images in the marketplace, tiers, backup, migration and studio groups have not been re-reviewed. Every one sampled showed pre-redesign chrome.
• Repo-wide, roughly 898 alt-text tags containing & render double-escaped. Pre-existing pipeline behaviour, deliberately out of scope here.