PostgreSQL ODBC Driver hangs during SQLExecDirect

Question

This happens just in prod environment and I could not reproduce same behaviour in a sandbox.

#0  0x00007f07854d8cb9 in __GI___poll (fds=0x7f076d8f1e88, nfds=1, timeout=-1) at ../sysdeps/unix/sysv/linux/poll.c:29
#1  0x00007f077c2ad990 in ?? () from /usr/lib/x86_64-linux-gnu/libpq.so.5
#2  0x00007f077c2adaf8 in ?? () from /usr/lib/x86_64-linux-gnu/libpq.so.5
#3  0x00007f077c2aa2d9 in PQgetResult () from /usr/lib/x86_64-linux-gnu/libpq.so.5
#4  0x00007f077c5176b3 in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#5  0x00007f077c4fa644 in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#6  0x00007f077c4fbab5 in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#7  0x00007f077c53242a in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#8  0x00007f077c50de7a in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#9  0x00007f077c50f3a7 in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#10 0x00007f077c50fac1 in ?? () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#11 0x00007f077c5395bb in SQLExecDirect () from /usr/lib/x86_64-linux-gnu/odbc/psqlodbcw.so
#12 0x00007f0787a5a16f in SQLExecDirect () from /usr/lib/x86_64-linux-gnu/libodbc.so.2
#13 0x00000000004be434 in my_function_where_i_call_sql_exec_direct ()

We use managed Postgres in a cloud similar to AWS, this service provides web interface where I have an access to server logs. The most strange thing to me that I see my query successfully executed:

{
    "transaction_id": "0",
    "error_severity": "LOG",
    "hostname": "rc1a-...mdb.yandexcloud.net",
    "internal_query": "",
    "process_id": "2624784",
    "session_id": "60672f13.280d10",
    "query_pos": "0",
    "user_name": "ace",
    "application_name": "ace-realm-meta - 10.128.0.9 ",
    "command_tag": "idle in transaction",
    "context": "",
    "hint": "",
    "message": "statement: RELEASE _EXEC_SVP_0x319d9f0;SAVEPOINT _EXEC_SVP_0x319d9f0;SELECT R_Object.meta FROM R_Object  WHERE ((R_Object.id \\= 15) ) ORDER BY R_Object.id ASC ",
    "session_line_num": "2817",
    "connection_from": "localhost:48058",
    "database_name": "droblozhko1",
    "internal_query_pos": "0",
    "query": "",
    "virtual_transaction_id": "59/191031",
    "detail": "",
    "location": "",
    "session_start_time": "2021-04-02T17:49:55+03:00",
    "sql_state_code": "00000"
}

(which means no exclusive locks are blocking my select?!)

I have googled a lot and found just one discussion about similar story: https://postgrespro.com/list/thread-id/2394695, but SQL_ATTR_QUERY_TIMEOUT nor keepliave pqoptions mentioned in this thread are affecting anything :\

Please, share any thoughts because I'm completely run out of ideas.

Update1. Why timeout is negative(in a last frame)? This is why poll() never returns.

Update2 I found out that PQgetResult() calls for pqWait() which calls pqWaitTimed with hardcoded finish_time to -1. The question now is how this logic supposed to deal with network issues?

PostgreSQL ODBC Driver hangs during SQLExecDirect

0 Answers0