Disconnection issues on both Linux & Windows VPS

[Jaddevvv]

[Linux Docker 5.9.11] cTrader CLI gets stuck in Login failed reconnect loop

Summary

A cBot running through the official Linux Docker image loses its cTrader connection. The container and cTrader processes remain alive, but the CLI cannot complete account re-authentication.

After several failed login retries, the CLI reports that the connection was restored, but no subsequent Logged in. message appears. The cBot then stops producing normal market-processing logs.

A separate Windows VPS running cTrader CLI version 5.7.14.51420 successfully reconnects and logs in after similar connection interruptions.

Linux Docker environment

• Image: ghcr.io/spotware/ctrader-console:latest
• CLI version: 5.9.11.0
• Image digest: <hash>
• Linux VPS running Docker
• Container: <redacted-container-name>
• cBot: <CbotName>
• Symbol: <Symbol>
• Period: <Timeframe>
• Broker/account details: redacted
• The VPS SSH connection remained available during the failures.

The cBot was started with the equivalent of:

ctrader-cli run <algo-file> <cbotset-file> \
  --ctid=<redacted> \
  --pwd-file=<redacted> \
  --account=<redacted> \
  --broker=<redacted> \
  --symbol=<Symbol> \
  --period=<Timeframe> \
  --exit-on-stop

Docker state during the problem:

Status=running
Running=true
RestartCount=0
OOMKilled=false

The cTrader CLI process and algohost.netcore process were still running.

Linux startup

The cBot initially started and authenticated successfully:

The connection has been established.
Login to <redacted-account>...
Logged in.
Starting cBot...
Info | CBot instance [XXX] started.

The bot then produced normal trading activity, including successful market orders, position closes and stop-loss placement.

Linux successful reconnects

Some earlier interruptions recovered successfully:

10/09/2026 13:34:55 | The connection has been lost. Reconnecting...
10/09/2026 13:35:04 | Establishing connection...
10/09/2026 13:35:05 | The connection has been restored.
Login to <redacted-account>...
Logged in.

Linux failed reconnect loop

On 14 September 2026, the CLI entered this sequence:

2026-09-14T13:26:25.788Z Establishing connection...
2026-09-14T13:26:26.384Z The connection has been restored.
2026-09-14T13:26:26.480Z Login to <redacted-account>...
2026-09-14T13:26:26.480Z Connection Error: Unknown
2026-09-14T13:26:26.480Z Login failed. Retrying in 1 seconds...
2026-09-14T13:26:27.484Z Login to <redacted-account>...
2026-09-14T13:26:27.484Z Connection Error: Unknown
2026-09-14T13:26:27.484Z Login failed. Retrying in 2 seconds...
2026-09-14T13:26:29.488Z Login to <redacted-account>...
2026-09-14T13:26:29.488Z Connection Error: Unknown
2026-09-14T13:26:29.488Z Login failed. Retrying in 4 seconds...
2026-09-14T13:26:33.492Z Login to <redacted-account>...
2026-09-14T13:26:33.492Z Connection Error: Unknown
2026-09-14T13:26:33.492Z All login retry attempts failed, moving to reconnection state.
2026-09-14T13:26:33.492Z The connection has been lost. Reconnecting...
2026-09-14T13:26:33.492Z Establishing connection...
2026-09-14T13:26:33.540Z The connection has been restored.

After the final The connection has been restored message, no later Logged in. message appeared.

The container remained running and did not restart automatically. A separate read-only cbots query successfully authenticated to the account but returned:

{
  "cbots": [],
  "count": 0
}

Separate Windows comparison — cTrader CLI 5.7.14.51420

A separate Windows VPS, using a different provider and network, runs cTrader CLI version 5.7.14.51420:

PS C:\Users\Administrator> ctrader-cli.exe --version
Version: 5.7.14.51420

The Windows cBot was started using the equivalent command:

ctrader-cli.exe run "<redacted-algo-file>.algo" \
  --ctid=<redacted> \
  --pwd-file=<redacted> \
  --account=<redacted> \
  --symbol=<Symbol> \
  --period=<Timeframe> \
  --environment-variables \
  --full-access

The Windows cBot started normally:

The connection has been established.
Login to <redacted-account>...
Logged in.
Starting cBot...
Info | CBot instance [XXXX] started.
Info | <cBot> started.
Info | Connected: True

Windows reconnect 1

12/09/2026 20:37:39.338 | The connection has been lost. Reconnecting...
12/09/2026 20:37:40.594 | Establishing connection...
12/09/2026 20:37:40.884 | The connection has been restored.
Login to <redacted-account>...
Logged in.
Info | <cBot> CONNECTION RESTORED

Windows reconnect 2

13/09/2026 07:03:40.948 | Info | <cBot> CONNECTION LOST
13/09/2026 07:03:42.173 | The connection has been lost. Reconnecting...
13/09/2026 07:03:44.643 | Establishing connection...
13/09/2026 07:03:44.945 | The connection has been restored.
Login to <redacted-account>...
Logged in.
Info | <cBot> CONNECTION RESTORED

The Windows process successfully re-authenticated after both interruptions without a container or server restart.

Comparison and suspected scope

Both environments experience intermittent cTrader connection interruptions.

The Windows CLI successfully completes the reconnect and account login. The Linux Docker CLI reaches The connection has been restored, then enters repeated Connection Error: Unknown and Login failed messages.

This suggests that the failure may be specific to the Linux Docker reconnect/authentication path or to a version-specific regression. The Windows and Linux CLI versions are different, so this comparison is not intended to claim that both builds use identical code.

Docker sees a running process even though the trading session may no longer be authenticated.

Questions

1. Is this a known reconnect/authentication issue in cTrader Console 5.9.11?
2. Why does the CLI report The connection has been restored before account login still fails?
3. Should the CLI recreate the account authentication/session after a socket reconnect?
4. Should the CLI exit with a non-zero code after all login retries fail?
5. Is there a recommended workaround other than manually restarting the container?
6. Is a healthcheck or authenticated-session status planned for unattended Docker deployments?
7. Is the reconnect behavior expected to differ between the Windows CLI and Linux Docker CLI?
8. Is there a known-good CLI version that avoids this Linux/Windows reconnect loop?

[amusleh-spotware-com]

Thanks for the detailed report. We reproduced part of it, and it uncovered two more bugs.

Confirmed

The CLI never exits, regardless of outage length. We tested ghcr.io/spotware/ctrader-console:5.9.11 with the network cut for 90 seconds, 20 minutes, and five 2-minute cycles. The container stayed Running=true, Exit=0, RestartCount=0 in every case. The 30-second connection watchdog is only started for the accounts and symbols commands, not for run, so run has no timeout, no failure exit, and --restart never fires.

Connection Error: Unknown is not a server or network error. It is an internal placeholder returned when the connection state machine is asked to log in from a state that cannot perform a login. That is why it appears in the same millisecond as the Login to ... line. Retrying it cannot succeed.

Not reproduced

Network loss alone never killed the session in our tests — all four outage patterns reconnected cleanly. The CLI only enters that Login failed loop when auto-reconnect gives up, and it gives up only on a server-side rejection (expired or invalid trading account token, account not found), never on connectivity loss. Your log shows you reached that loop, so the re-authentication was rejected server side. We cannot trigger that from the client.

One clarification

Logged in. is not reprinted on a reconnect. A healthy reconnect prints The connection has been lost... then The connection has been restored. and nothing more, because the re-login is silent. Its absence is not a symptom, and neither is a quiet cBot if it only logs on trades.

This is also not Linux or Docker specific. The reconnect path is shared, OS-agnostic code, so 5.7.14 has the same defects. There is no known-good version — 5.7, 5.9 and 5.10 are all affected.

Fixes

1. --reconnect-timeout=<seconds> for run. If the session is not restored within the budget, the CLI exits non-zero with a reason, so --restart works. Opt-in; omitting it keeps current behaviour.
2. Unknown no longer triggers three retries. It goes straight to a full reconnect.
3. The account list load now retries with backoff (1s to 1min) instead of failing once, silently.
4. ctrader-logs=1 alone enables logging. It previously also required an undocumented ctrader-test-tools=1.

These are on the 5.10 branch. We will comment here when the image is published.

If it happens again

Please capture the container log with both -e ctrader-logs=1 -e ctrader-test-tools=1 (both are required on the current image). We need the ConnectionStateMachine: entering state ... lines.

Note: the diagnostic log currently records the cTrader ID password in clear text. That is our bug and we are fixing it separately. Until then, remove the Password= value under ProtoCHAuthCtidReq before sharing a log.

Your questions

1. Partly. A different reconnect bug was fixed in 5.9.10/5.9.11. Yours is new and now tracked.
2. That line reports the cTrader ID connection, not the trading account session. We will reword it.
3. It does try — see the Unknown explanation above.
4. Yes, fix 1.
5. No clean workaround today. Log a heartbeat from the cBot and restart the container externally if it goes silent.
6. Not planned before this. We are raising it — a healthcheck should verify the authenticated session, not just the process.
7. No. Same code on both. The difference is the command lines and the networks.
8. No.

[Jaddevvv]

Thanks, I'm always running Ctrader-cli instances on various VPS using both linux and windows for myself and for my clients. That will help a lot when it will be fixed.

If you need anything from me besides the logs, do not hesitate to let me know 😁 !