fix(opcuaclient): split the per-endpoint connect budget into reachability + session phases (#912)

PerEndpointConnectTimeout (3 s) now bounds only TCP + discovery, so failover past a
dead endpoint stays fast. The new SessionCreateTimeout (10 s) bounds OpenSecureChannel
+ CreateSession on an endpoint that answered discovery, where the client-side cert and
crypto work lives (#910). 2 x (3 + 10) s stays inside the host's 30 s initialise bound.
Expiries name the budget that ran out; both budgets must be positive; AdminUI form
carries the new field so a save does not drop it; driver docs updated.

Closes #912
This commit is contained in:
Joseph Doherty
2026-11-15 16:55:22 -05:00
parent 8ff424fc5a
commit 678498a4a6
6 changed files with 150 additions and 29 deletions
+10 -4
View File
@@ -75,10 +75,16 @@ measurements; they are not fresh SDK 2.0.1 interoperability evidence.
`EndpointUrls` (a list) takes precedence over the single-URL `EndpointUrl` and
provides ordered **failover** — the driver tries each candidate in turn at every
`InitializeAsync` (including the comms-loss watchdog's re-initialise), and the first to connect wins (e.g. a hot-standby pair on
4840 / 4841). Each candidate gets `PerEndpointConnectTimeout` (default 3 s) for discovery + secure channel + session. That
budget includes **client-side** work (certificate load/validation, the asymmetric handshake), so a cold process on a busy
host can exceed it against a healthy server; an expiry is reported as `TimeoutException: … PerEndpointConnectTimeout (3 s)`
(#910 — the SDK alone says `BadRequestInterrupted`, which reads like a server fault). See
4840 / 4841). Each candidate's connect runs in **two separately bounded phases** (#912):
`PerEndpointConnectTimeout` (default **3 s**) covers TCP + endpoint discovery, so a dead or unreachable candidate is
passed over in about 3 s; `SessionCreateTimeout` (default **10 s**) then covers OpenSecureChannel + CreateSession +
ActivateSession on an endpoint that answered discovery. The session phase carries the **client-side** work (certificate
load/validation, the asymmetric crypto of a secured handshake), which a cold process on a busy host can stretch past 3 s
against a healthy server (#910). The larger budget only costs time when an endpoint answers discovery and then stalls —
not when it is down. The defaults keep two such candidates (2 × (3 + 10) s) inside the host's 30 s initialise bound;
raise either per driver if your topology needs it, keeping that sum in mind. An expiry is reported by the budget's name,
e.g. `TimeoutException: … SessionCreateTimeout (10 s)` (#910 — the SDK alone says `BadRequestInterrupted`, which reads
like a server fault). The AdminUI browse picker still uses one combined budget (`PerEndpointConnectTimeout` clamped to 5–30 s). See
`src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.OpcUaClient.Contracts/OpcUaClientDriverOptions.cs`
for every field (security policy/mode, auth type, session timeout, keep-alive,
browse root, node/depth caps). `ReconnectPeriod` was **deleted** (#676 — unused since #670 removed the SDK reconnect