Session restoration fails permanently due to gopcua CloseSession during reconnect

[skartikey]

Relevant telegraf.conf

[[inputs.opcua_listener]]
    endpoint = "opc.tcp://<server>:4863"
    security_policy = "None"
    security_mode = "None"
    # subscription with 78 monitored items

Logs from Telegraf

debug: client: monitor: disconnected
  debug: client: monitor: auto-reconnecting
  debug: client: monitor: action: createSecureChannel
  debug: client: monitor: secure channel recreated   
  debug: client: monitor: action: restoreSession  
  debug: client: monitor: trying to restore session
  # ActivateSession succeeds, then CloseSession is sent immediately
  uasc 3/2: recv *ua.ActivateSessionResponse                       
  uasc 3/3: send *ua.CloseSessionRequest    
  uasc 3/3: recv *ua.CloseSessionResponse
  debug: client: monitor: session restored
  debug: client: monitor: trying to update namespaces
  uasc 3/4: err: The session id is not valid. StatusBadSessionIDInvalid (0x80250000)
  debug: client: monitor: updating namespaces failed: The session id is not valid.  
  # Second attempt: recreateSession succeeds but TransferSubscriptions fails      
  debug: client: monitor: session recreated                                 
  debug: client: monitor: action: transferSubscriptions
  uacp 4: recv ERRF with 16 bytes                      
  debug: client: monitor: transfer subscriptions failed. Recreating all subscriptions: EOF
  debug: client: monitor: action: restoreSubscriptions                                    
  debug: sub 11: recreate_create: failed to recreate subscription
  debug: client: monitor: recreate subscripitions failed: write tcp ...: broken pipe
  debug: client: monitor: no subscriptions to resume                                                                                                                                                               
  # Telegraf is now permanently disconnected, no further reconnection attempts

System info

Telegraf 1.38.2 (Docker), connecting to Siemens OpenPCS7 V9.2 OPC UA Server

Docker

No response

Steps to reproduce

1. Configure opcua_listener with subscription to a Siemens OpenPCS7 server
2. Wait for a period of low data change activity (15-25 minutes with few/no value changes)
3. The server drops the TCP connection (EOF)
4. gopcua's auto-reconnect triggers but fails permanently

Expected behavior

Telegraf should recover from connection drops and re-establish subscriptions, either through gopcua's auto-reconnect or through its own reconnection logic.

Actual behavior

After the connection drops, gopcua's restoreSession path calls ActivateSession which internally calls CloseSession on the same session (with DeleteSubscriptions: true). This destroys the session and its subscriptions. The subsequent TransferSubscriptions and restoreSubscriptions both fail because the server closes the connection. Telegraf ends up permanently disconnected with no further reconnection attempts.

Additional info

Root cause analysis:

The bug is in gopcua v0.8.0's ActivateSession method (client.go:906). During reconnection, when restoreSession calls ActivateSession(ctx, existingSession), the method's response handler calls c.CloseSession() on the "previous" session. But during restoreSession, the previous session IS the session being restored, so it immediately closes what it just activated.

This was partially fixed in gopcua/opcua#700 for the recreateSession path (by setting c.setSession(nil) before calling ActivateSession), but the restoreSession path was not fixed.

[skartikey]

Raised a PR in the upstream library (gopcua/opcua#855) to address the root cause.

The issue is in the restoreSession path of gopcua’s auto-reconnect logic. It calls ActivateSession, which internally invokes CloseSession on the same session it just restored (with DeleteSubscriptions: true). This ends up deleting all subscriptions and makes recovery impossible.

The fix is a simple one-line change, consistent with what was already done for the recreateSession path in gopcua/opcua#700.

Once this is merged and released, we can update the gopcua dependency in Telegraf.

In the meantime, restarting Telegraf is the only reliable workaround to re-establish the connection.

[sahil-stellar]

Hi @skartikey , I am also having a similar kind of error but slightly different. In my case, subscription was successfully registered but recreation had intermittent error due to OPC nodes becoming invalid and throwing StatusBadNodeIDUnknown errors. My hunch is, the session was closed due to intermittent network issue between OPC server and Telegraf machine, causing request_timeout to close the session. Just before this, all the nodes went StatusBad

debug: publish: notAcked=[]
debug: publish: notif: 31471
debug: publish: pendingAcks=[{"SubscriptionID":6647,"SequenceNumber":31471}]
debug: publish: PublishRequest: {"RequestHeader":null,"SubscriptionAcknowledgements":[{"SubscriptionID":6647,"SequenceNumber":31471}]}
debug: publish: PublishResponse: null
debug: publish: error: unexpected error. Do we need to stop the publish loop?: The session was closed by the client. StatusBadSessionClosed (0x80260000)
debug: sub: error: The session was closed by the client. StatusBadSessionClosed (0x80260000)
debug: sub: pause
debug: client: monitor: disconnected
debug: client: monitor: auto-reconnecting
debug: client: monitor: action: createSecureChannel
debug: sub: pause: pause
debug: client: monitor: trying to recreate secure channel
debug: client: monitor: secure channel recreated
debug: client: monitor: action: restoreSession
debug: client: monitor: trying to restore session
debug: client: monitor: restore session failed: The session id is not valid. StatusBadSessionIDInvalid (0x80250000)
debug: client: monitor: action: recreateSession
debug: client: monitor: trying to recreate session
debug: client: monitor: session recreated
debug: client: monitor: trying to update namespaces
debug: client: monitor: namespaces updated
debug: client: monitor: action: transferSubscriptions
debug: client: monitor: sub 6647: transfer subscription failed
debug: client: monitor: action: restoreSubscriptions
debug: sub 6647: recreate_delete: subscription deleted
debug: sub: pause: pause
debug: sub 6647: recreate_create: recreated as subscription 8581
sub 8581: recreate: subscription registered
debug: client: monitor: recreate subscripitions failed: The node id refers to a node that does not exist in the server address space. StatusBadNodeIDUnknown (0x80340000)
debug: client: monitor: no subscriptions to resume

This also caused Telegraf to permanently disconnect. Will both of these similar but distinct issues be fixed with (gopcua/opcua#855)?

If not, is there something that can be implemented in the telegraf source code for a quick fix?

Much thanks

[skartikey]

@sahil-stellar This is a different root cause, so #855 won't fix yours. Please open a new issue with full config, gopcua/Telegraf versions, and debug logs so we can track it separately and file it upstream.