Skip to main content
Balancing the inputs of a regular Flow and balancing the transit pool operate at different sites, and switching the route through the Flow Guard is another, separate operation: it changes the applied configuration after a confirmed failure.

Where the policy is set

Pool strategies

The policy is executed locally on each server that selects the next transition. For the first transition, the pool is the previous node of a regular Flow; for the next one, the corresponding transit node. The interval, timeout, check address, sample, expected number of candidates, and fallback parameters affect the selection. Use the help at the field: the set of settings depends on the strategy.

Refusal behaviour

Inaccessibility is accounted for after observation by a breakdown, so the switch is not instantaneous. Existing connections to a fallen server can be terminated: reserve does not guarantee the preservation of the session. The standard fail_closed pool policy blocks traffic when there are no available candidates. Direct bypass should not be considered an implicit reserve. The Flow Guard can separately select the priority reserve, the RTT from the CTRL or the CPU agent. These are different dimensions and a different mechanism than the leastLoad strategy in the transit pool.

Verification

After application, check the distribution of new connections and the availability of each transition, check the behavior of the loss of one and all candidates in a dedicated test environment, and the Failure Lab calculation does not replace such a run.

Rollback

Return the parameters in the draft and re-apply the affected Flow or restore the working version. Changing the policy of the pool directory does not change the images already used. Check the availability of features in the installed version in the panel; this reconciliation does not confirm the deployment on the servers.