Keeping sets working

A strategy that works today can stop working later. The watchdog checks a set's pages on a schedule. When they keep failing, it heals the set: it searches for a new strategy with Discovery and writes it into the set. Escalation adds a second safety net: it moves a site that keeps failing to a backup set for a while, then tries the original set again.

Requirements

• A set that applies a bypass strategy. A set that routes its traffic to a proxy or an interface cannot be watched.
• The set targets domains or GeoSite categories, and is not limited to specific devices by an include list.
• Up to five pages that represent the set and do not open without b4, such as the main page of each service it covers.

Setup

1. Open Settings > Discovery, turn on Enable Watchdog and click Save Changes.
2. Open Sets and open the set.
3. On the Discovery tab, enter a page under Address and press Enter. Repeat for the other pages. Suggest offers candidates, and Add puts one in the list.
4. Turn on Keep this set working with the watchdog.
5. Click Save Changes.

A backup set for escalation is optional:

1. On the set's Escalation tab, pick the backup set under Switch to. A set that sends sites through a VPN or a proxy can be the backup.
2. Click Save Changes.

The fields of these tabs are described in Discovery tab and Escalation.

Checking the result

The Watchdog page lists the set under Watched sets. When the first check passes, its status reads Healthy. The set's card on the Sets page carries a Watchdog badge, and the dashboard's Sets panel shows the same status.

When the watchdog writes a new strategy, the row shows it under Last heal, and the dashboard lists it under Recent changes. While escalation sends a site to the backup set, the dashboard's Active Escalations panel lists it.

When it does not work

Status or message	Cause and fix
The watchdog is off, but N sets ask for it, on the dashboard	Enable Watchdog is off under Settings > Discovery. Turning it on starts the checks.
The switch cannot be turned on	A condition fails, and the caption under the switch names it, see Watchdog.
Cannot verify	No page can be checked: another set handles them, an escalation sends them elsewhere, or they are unusable. The set needs pages that it handles itself.
Cooldown or Gave up, with the reason that the current strategy loads the addresses	Tested alone, the set's strategy loads the pages, yet the live check fails. Something outside the strategy is in the way: another set, DNS or IPv6. The Traffic page shows which set the connections get.
Gave up	Three heals in a row failed, and the reason shows the last cause. A manual run with Find a strategy on the set's Discovery tab shows what works, and Check now on the Watchdog page resumes the heals.

When no strategy helps any more, the sites need a VPN or a proxy. A set routed that way, as the escalation backup, catches the sites that keep failing.