One failed query can prevent future queries working

[seanhinde]

In the lab Oracle right click and "Browse Data" on the table SCOTT.NUMBER_LONG. The erlang console shows the error message:

(dderl1@127.0.0.1)1> {oci_logger,108} TCP closed
{oci_logger,137} Terminating normal
16:23:48.389 [error] [_DDRL_] {c_src/erloci_lib/ocistmt.cpp:execute:707} Unsupported column type 8
16:23:48.389 [info] [_DDRL_] {dderloci,358} RowIdError {error,{0,<<"[execute:706] unsupporetd type 8\n">>}}
16:23:48.404 [error] [_DDRL_] {dderloci,59} run_query(select * from SCOTT.NUMBER_LONG,undefined,select SCOTT.NUMBER_LONG.*, SCOTT.NUMBER_LONG.ROWID from SCOTT.NUMBER_LONG)
{normal,{gen_server,call,[<0.834.0>,{port_call,[8,140339042098512,140339045211840]},infinity]}}
16:23:48.404 [error] [_DDRL_] {oci_adapter,693} query error {error,normal}

Further attempts to view previously working tables result in noproc errors e.g.

(dderl1@127.0.0.1)1> 16:24:16.924 [error] [_DDRL_] {dderloci,59} run_query(select * from SCOTT.TEST_TYPES,undefined,select SCOTT.TEST_TYPES.*, SCOTT.TEST_TYPES.ROWID from SCOTT.TEST_TYPES)
{noproc,{gen_server,call,[<0.834.0>,{port_call,[4,140339042098512,<<"select SCOTT.TEST_TYPES.*, SCOTT.TEST_TYPES.ROWID from SCOTT.TEST_TYPES">>]},infinity]}}
16:24:16.924 [error] [_DDRL_] {oci_adapter,693} query error {error,noproc}

Logging out and back in gets things back to a working state.

[c-bik]

I was under the impression that DDErl can gracefully handle all query errors (without loosing the Oracle conection session). @acautin can you please investigate why isn't that the case here?

[acautin]

@c-bik from:

{oci_logger,108} TCP closed
{oci_logger,137} Terminating normal
16:23:48.389 [error] [_DDRL_] {c_src/erloci_lib/ocistmt.cpp:execute:707} Unsupported column type 8
16:23:48.389 [info] [_DDRL_] {dderloci,358} RowIdError {error,{0,<<"[execute:706] unsupporetd type 8\n">>}}

Doesn't this means the connection was terminated on the driver side ?

[c-bik]

@acautin
Yes if...

{oci_logger,108} TCP closed
{oci_logger,137} Terminating normal

...and...

[error] [_DDRL_] {c_src/erloci_lib/ocistmt.cpp:execute:707} Unsupported column type 8
[info] [_DDRL_] {dderloci,358} RowIdError {error,{0,<<"[execute:706] unsupporetd type 8\n">>}}

...are related and for same cause...
However c_src/erloci_lib/ocistmt.cpp:execute:707} +/- few lines

            default:
				r.fn_ret = FAILURE;
				SPRINT(r.gerrbuf, sizeof(r.gerrbuf), "[%s:%d] unsupporetd type %u\n", __FUNCTION__, __LINE__, cur_clm.dtype);
				REMOTE_LOG(ERR, "Unsupported column type %d\n", cur_clm.dtype);
				throw r;
                break;

is no different from any other SQL execute error including table doesn't exist (in which case I assume DDErl doesn't, and it shouldn't, loose the session).

[acautin]

dderl doesn't, isn't the driver the one that is closing the connection ?, that was my question as AFAIK we don't try to understand the errors just forward them to the client.

[c-bik]

if that is the case than, specifically,

select SCOTT.NUMBER_LONG.*, SCOTT.NUMBER_LONG.ROWID from SCOTT.NUMBER_LONG

is causing the connection close!

@seanhinde can you please try the following two queries from erloci level (without DDErl) and post the console logs here?

select * from SCOTT.TABLE_DOESNT_EXIST

and

select SCOTT.NUMBER_LONG.*, SCOTT.NUMBER_LONG.ROWID from SCOTT.NUMBER_LONG

[acautin]

reported to erloci K2InformaticsGmbH/erloci#71