Recover local Sendico session after BFF network restarts (#57)
This commit is contained in:
1 parent
021bd04324
commit
13cc9c302e
9 files changed
+158
-8
No files matched your search
@@ -105,6 +105,27 @@ docker compose --project-directory . -f infra/local/compose.yml \
|
||||
-f infra/local/sendico/compose.yml --env-file .local/dev/images.env restart sendico
|
||||
```
|
||||
|
||||
A BFF container restart can leave its session worker in the old shared network
|
||||
namespace. The worker checks Vault DNS every ten seconds, with a five-second
|
||||
lookup limit. Three consecutive failures log a credential-free recovery reason
|
||||
and terminate the session process; Docker's restart policy then attaches it to
|
||||
the BFF's current namespace. A successful lookup resets the counter, so transient
|
||||
DNS failures do not immediately restart the worker.
|
||||
|
||||
Container recreation changes the BFF container ID as well as its namespace.
|
||||
Bootstrap and credential-rotation tooling recreate BFF and its configured session
|
||||
worker together. Compose also marks the session dependency for restart when BFF
|
||||
is explicitly updated. For a manual forced recreation, select both services:
|
||||
|
||||
```sh
|
||||
docker compose --project-directory . -f infra/local/compose.yml \
|
||||
-f infra/local/sendico/compose.yml --env-file .local/dev/images.env \
|
||||
up -d --no-deps --force-recreate --wait bff sendico
|
||||
```
|
||||
|
||||
The parent process cleans up its bridge, DNS monitor and renewal delay on exit.
|
||||
This recovery retains the configured provider URL, account and replay namespace.
|
||||
|
||||
`make test-dev` checks seven healthy services, verified browser TLS, the Vault
|
||||
layout and least-privilege identities, Mongo transactions/lease rotation, and
|
||||
authenticated Sendico currency and ledger reads through the same loopback
|
||||
@@ -130,6 +151,21 @@ and the established CA. A subsequent verified TLS check must succeed.
|
||||
|
||||
## Acceptance and operator UX review
|
||||
|
||||
Recovery follow-up for [task #57](https://git.sendico.io/tech/sxchnge/issues/57) /
|
||||
[PR #59](https://git.sendico.io/tech/sxchnge/pulls/59), validated on 2026-10-19:
|
||||
`make test-dev-tooling`, `make test-dev-rotation`, Compose configuration validation,
|
||||
`sh infra/local/tools.sh verify` and `git diff --check` pass. A BFF-only restart
|
||||
triggered automatic session restart and restored matching network namespaces;
|
||||
paired recreation also finished healthy. Rotation observed three real Vault
|
||||
pool replacements, an unchanged BFF process and 534 transaction-backed requests
|
||||
with zero failures. The final check verifies all seven SXCHNGE services, browser
|
||||
HTTPS, Vault/database boundaries and authenticated Sendico currency/ledger reads.
|
||||
An earlier check failed because Sendico's ledger announcement connection was
|
||||
closed; restoring Discovery and its service announcers in dependency order
|
||||
recovered that existing stand. No Sendico source or financial data was changed,
|
||||
and no email or payment was sent. This follow-up changes shell/Compose tooling;
|
||||
it does not rerun the earlier full Go/Flutter suite for unchanged product source.
|
||||
|
||||
Validated on 2026-10-16: `make build`, `make up`, `make test-dev-tooling`,
|
||||
`make test-dev-infra`, `make test-dev`, and `git diff --check` passed. Both product
|
||||
sites, Sendico Backoffice and both API health endpoints returned HTTP 200;
|
||||
|
||||
Reference in new issue
Block a user