mongodb-js / devtools-shared / 26892795218 / 9 – mongodb-cloud-info

79%

main: 79%

push

github

fix(devtools-proxy-support): reconnect SSH tunnel after unrecoverable client error COMPASS-8355 (#772)

* fix(devtools-proxy-support): reconnect SSH tunnel after unrecoverable client error COMPASS-8355

- Add sshClient.on('error') handler in SSHAgent constructor to prevent
  unhandled 'error' events from crashing the process when the SSH session
  dies unexpectedly (e.g. after hibernate)
- Add reinitializeClient flag to initialize() to recreate the ssh2 Client
  instance when it enters an unrecoverable "Instance unusable" state
- Expand retryable error patterns to include "Instance unusable after fatal
  error", "read ECONNRESET", and "Socket closed"
- Extract client setup into createSshClient() to ensure close/error handlers
  and forwardOut binding are consistent on both initial setup and recreation

* address review feedback: use = undefined, improve test coverage

- Replace delete this.connectingPromise with = undefined assignment
- Improve "Instance unusable" test: patch connect() to simulate broken
  client state, verify fresh client is created and auth happens twice
- Track initializedSuccessfully in _connect() to detect when initialize()
  itself failed, requiring a new client on retry

* fix(devtools-proxy-support): remove unused once import COMPASS-8355

* fix(devtools-proxy-support): restore end() call in reinitializeClient to clear keepalive timer COMPASS-8355

Without this, the old ssh2 Client's keepalive setInterval keeps firing
after initialize(true) replaces it, preventing process exit. ssh2's end()
guards with isWritable() so it is a no-op when the socket is already dead.

* test(devtools-proxy-support): verify SSH reconnect after server-side connection reset COMPASS-8355

11 of 11 branches covered (100.0%)

Branch coverage included in aggregate %.

32 of 34 relevant lines covered (94.12%)

38940.44 hits per line