Hi, We have a Python application that spawns multiple threads the write the the database. We are seeing an intermittent problem where one of the threads will "block" seemingly for ever. We saw a similar problem twice with the same signature in a C++ based program as well (but that seems harder to reproduce). This is PostgreSQL 8.1.8-1 on a Fedora Core 5 system (x86_64 arch). The PostgreSQL library used is the Python pgdb library. The The backtrace for the client program is: #0 0x000000321e6c3086 in poll () from /lib64/libc.so.6 #1 0x0000003379c0f775 in pqSocketCheck (conn=0x61f070, forRead=1, forWrite=0, end_time=-1) at fe-misc.c:1039 #2 0x0000003379c0f870 in pqWaitTimed (forRead=Variable "forRead" is not available. ) at fe-misc.c:913 #3 0x0000003379c0e792 in PQgetResult (conn=0x61f070) at fe-exec.c: 1186 #4 0x0000003379c0e86e in PQexecFinish (conn=0x61f070) at fe-exec.c: 1415 #5 0x00002aaaae91b341 in pgsource_execute (self=0x2aaaaea4da50, args=Variable "args" is not available. ) at pgmodule.c:534 #6 0x00000032283954a0 in PyEval_EvalFrame () from /usr/lib64/ libpython2.4.so.1.0 The process is connected to the database on the local machine. The connection information in ps says "idle in transaction". The backtrace for the postgresql backend process at that time: (gdb) bt #0 0x000000321e6cc5e5 in recv () from /lib64/libc.so.6 #1 0x00000000005004f6 in secure_read (port=0xa0e0f0, ptr=0x7dc8c0, len=8192) at /usr/include/bits/socket2.h:35 #2 0x0000000000506114 in pq_recvbuf () at pqcomm.c:697 #3 0x0000000000506537 in pq_getbyte () at pqcomm.c:738 #4 0x00000000005688b7 in PostgresMain (argc=4, argv=0x9ef7d8, username=0x9ef7a0 "user") at postgres.c:289 #5 0x00000000005422fb in ServerLoop () at postmaster.c:2865 #6 0x0000000000543234 in PostmasterMain (argc=5, argv=0x9ec590) at postmaster.c:941 #7 0x0000000000507ebe in main (argc=5, argv=0x1a) at main.c:265 There does not appear to be any deadlock: select * from pg_locks; locktype | database | relation | page | tuple | transactionid | classid | objid | objsubid | transaction | pid | mode | granted -----------+----------+----------+------+------- +---------------+---------+-------+----------+-------------+------- +-----------------+---------nsactionid | | | | | 74260015 | | | | 74260015 | 16855 | ExclusiveLock | t relation | 217529 | 1247 | | | | | | | 74260015 | 16855 | AccessShareLock | t transactionid | | | | | 74531807 | | | | 74531807 | 10377 | ExclusiveLock | t relation | 217529 | 10342 | | | | | | | 74695161 | 22791 | AccessShareLock | t transactionid | | | | | 74695161 | | | | 74695161 | 22791 | ExclusiveLock | t relation | 217529 | 218724 | | | | | | | 74260015 | 16855 | AccessShareLock | t (6 rows) \q libpq is built with thread-safety as far as I can tell - #define ENABLE_THREAD_SAFETY 1 in /usr/include/pg_config_x86_64.h Any suggestions for what else I can try to narrow down the problem? I found a couple of similar problems on the archives, but no solution that I could apply. Please let me know if there is any other information that can be collected to help debug this and I will collect it the next time the problem is see. Any suggestions appreciated. Thanks in advance, Venkat
