NJS-103: unexpected message type 1 received at position 54 of packet 8372

[damianhorex]

1. What versions are you using?

   Oracle 19c Enterprise edition version 19.21.0.0.0

   process.platform: linux
   process.version: v20.11.0
   process.arch: x64
   require('oracledb').versionString: 6.4.0
   require('oracledb').oracleClientVersionString: undefined

3. Is it an error or a hang or a crash?

   Error

4. What error(s) or behavior you are seeing?

   NJS-103: unexpected message type 1 received at position 54 of packet 8372
   at Object.throwErr (/node_modules/oracledb/lib/errors.js:617:10)
   at ExecuteMessage.processMessage (/node_modules/oracledb/lib/thin/protocol/messages/base.js:216:14)
   at ExecuteMessage.processMessage (/node_modules/oracledb/lib/thin/protocol/messages/withData.js:91:13)
   at ExecuteMessage.process (/node_modules/oracledb/lib/thin/protocol/messages/base.js:195:12)
   at ExecuteMessage.decode (/node_modules/oracledb/lib/thin/protocol/messages/base.js:186:10)
   at Protocol._decodeMessage (/node_modules/oracledb/lib/thin/protocol/protocol.js:78:17)
   at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
   at async Protocol._processMessage (/node_modules/oracledb/lib/thin/protocol/protocol.js:153:9)
   at async ThinConnectionImpl._execute (/node_modules/oracledb/lib/thin/connection.js:194:7)
   at async ThinConnectionImpl.execute (/node_modules/oracledb/lib/thin/connection.js:907:14)

I am not entirely confident there is no user error playing a role. The original system was written using oracledb v1.12.2, and we are currently working on moving to v6.4. It is entirely possible I am missing something, but was hoping posting this might provide some insight. The NODE_ORACLEDB_DEBUG_PACKETS output is attached. Thank you.

oracledb_debug_pkt_8372.txt

[sharadraju]

@damianhorex Thanks for using node-oracledb!

To debug this issue, please provide the following:

• A smaller reproducible test case
• Is this issue happening consistently?
• Have you tried this with a slightly earlier release like node-oracledb 6.2 or 6.3

This will help us in narrowing down the root cause.

Also, I am unable to download the oracledb_debug_pkt_8372.txt file.

[sudarshan12s]

Yes some information on minimal program, especially the sql command issued which throws the error would help . I think the database used is Oracle 19c Enterprise edition version 19.21.0.0.0

[damianhorex]

• A smaller reproducible test case
  

I will work on producing a smaller test case but it may take some time. This issue has come up while testing a large amount of raw telemetry data and it is occurring deeper in the processing chain. I was hoping the dump provided would help. If I am reading things correctly, it seems that packet 8372 is a database response to an insert of image header data into our OCAMS_IMAGE_HEADER table? At the moment some inserts are succeeding, some are failing, so I am looking at isolating differences in the data itself. I currently do not have the specs to decode the response in packet 8372 which is one of the reasons I posted.
I did have a co-worker sanity check the download of the debug data, he mentioned it was not working for him yesterday, but did work this morning. Maybe try again? I will still attach again below.

• Is this issue happening consistently?
  

Yes, I have consistently replicated it using 6.2, 6.3, and 6.4

• Have you tried this with a slightly earlier release like node-oracledb 6.2 or 6.3
  

Yes, my latest tests have shown a similar problem

oracledb_debug_pkt_8372.txt

Thanks!

[sudarshan12s]

@damianhorex , Thanks for sharing your findings. What i am seeing is that it looks you have a insert DML with large number of inbinds and 5 outbinds (NUMBER + 4 RAW) . While processing the response for outbind values, we are seeing some difference with what client decodes it as .. Correct me if its not what your test is trying to do..

We will meanwhile try to reproduce it locally . The database version is 19.21.0.0.0 right?

[sudarshan12s]

Can you share the following information if possible.

1. Does this test work in thick mode?
2. The insert DML fails for first time?
3. Do you change the input bind types for this insert DML subsequent execution or the bind information is provided with types fixed..

[damianhorex]

Thank you for your replies,
Yes, the database version is 19.21.0.0.0.
I will work on getting you answers to your other questions.

[damianhorex]

@damianhorex , Thanks for sharing your findings. What i am seeing is that it looks you have a insert DML with large number of inbinds and 5 outbinds (NUMBER + 4 RAW) . While processing the response for outbind values, we are seeing some difference with what client decodes it as .. Correct me if its not what your test is trying to do..

This is correct in this instance.

1. Does this test work in [thick](https://node-oracledb.readthedocs.io/en/latest/user_guide/initialization.html#enabling-node-oracledb-thick-mode-on-linux-and-related-platforms) mode?

I just tested in thick mode and I no longer encountered the error.

Note: I am still working on putting together a small test case that replicates the issue but wanted to share the results of the test in thick mode.

[sudarshan12s]

It looks if we have INOUT bind (like :reserved1) and it appears in returning into clause , issue seems to happen. We are checking more details and update.

[anthony-tuininga]

Can you try using a differently named bind variable for the returning clause? You are using the same name for input as well as the output for the returning clause -- and that isn't well supported. If I recall correctly, even though thick mode "works" it doesn't actually modify the output value. Can you confirm?

[anthony-tuininga]

I just confirmed it. You can't use bind variables with the same names before the returning clause and after the returning clause. In thick mode the value returned is not updated. We will ensure this situation doesn't raise that strange exception but in the meantime you can simply use different bind variable names before and after the RETURNING keyword to avoid this issue.

[damianhorex]

Ok, will do. Thank you for your time.

[manibhushan]

We are experiencing the same error code on our project. This happens while reading more rows (fetchMoreRows) on the resultSet. Is there any workaround available for this?

[sudarshan12s]

We are experiencing the same error code on our project. This happens while reading more rows (fetchMoreRows) on the resultSet. Is there any workaround available for this?

Can you check if your sql has INOUT same bind variable name (below example has str_val) appearing in input and also in RETURNING INTO clause like this. If this is the case, you need to use different names..

"INSERT INTO nodb_dmlreturn (id, name) VALUES (:int_val, :str_val) RETURNING id, name into :rid, :str_val"; binds = { int_val: 6, rid: { type: oracledb.NUMBER, dir: oracledb.BIND_OUT }, str_val: { type: oracledb.STRING, val: "A different test string", dir: oracledb.BIND_INOUT, maxSize: 200 } };