On Sat, Feb 7, 2009 at 3:45 PM, Tatsuo Ishii <[email protected]> wrote:
> pgpool-II 2.2 beta2 released. Changes from beta1: Perhaps another but, not sure though. I was executing this from my laptop: /usr/lib/postgresql/8.3/bin/pgbench -h postgresql.domain.com -p 9999 -U pgpool2 -d bench_replication -c 10 -t 1000 The process was going fine, until I decided to set the log_connections directive in the postgresql.conf of the slave node to "all". I did the /etc/init.d/postgresql reload while the process was still active and this appeared in the syslog of the node running pgpool: pgsql1 pgpool: 2009-02-11 11:52:45 ERROR: pid 26122: pool_process_query: 1 th kind S does not match with master or majority connection kind C pgsql1 pgpool: 2009-02-11 11:52:45 ERROR: pid 26122: kind mismatch among backends. Possible last query was: "UPDATE branches SET bbalance = bb alance + -3820 WHERE bid = 1;" kind details are: 0[C] 1[S] pgsql1 pgpool: 2009-02-11 11:52:45 LOG: pid 26122: notice_backend_error: 1 fail over request from pid 26122 pgsql1 pgpool: 2009-02-11 11:52:45 LOG: pid 26093: starting degeneration. shutdown host pgsql2.freyatest.domain(5432) pgsql1 pgpool: 2009-02-11 11:52:45 LOG: pid 26093: execute command: /var/lib/postgresql/8.3/main/pgpool-failover 1 pgsql2.freyatest.domain 5 432 /var/lib/postgresql/8.3/main 0 0 pgsql1 pgpool: sh: /var/lib/postgresql/8.3/main/pgpool-failover: No such file or directory pgsql1 pgpool: 2009-02-11 11:52:45 LOG: pid 26093: failover_handler: set new master node: 0 pgsql1 pgpool: 2009-02-11 11:52:45 LOG: pid 26093: failover done. shutdown host pgsql2.freyatest.domain(5432) Keep in mind the failover scripts were not "in position" yet, but that is incidental. In the PostgreSQL log in the slave node I saw this: 2009-02-11 11:52:45 CET LOG: unexpected EOF on client connection 2009-02-11 11:52:45 CET LOG: could not receive data from client: Connection reset by peer 2009-02-11 11:52:45 CET LOG: unexpected EOF on client connection And in the client machine this appeared: client 8 executing \setrandom delta -5000 5000 client 8 sending BEGIN; client 5 receiving Client 5 aborted in state 11: ERROR: kind mismatch among backends. Possible last query was: "UPDATE branches SET bbalance = bbalance + -3820 WHERE bid = 1;" kind d message contents do not agree with length in message type "E" client 8 receiving client 8 sending UPDATE accounts SET abalance = abalance + 1870 WHERE aid = 47397; client 2 receiving Client 2 aborted in state 10. Probably the backend died while processing. client 6 receiving Client 6 aborted in state 10. Probably the backend died while processing. client 1 receiving Client 1 aborted in state 11. Probably the backend died while processing. client 4 receiving Client 4 aborted in state 10. Probably the backend died while processing. client 8 receiving Client 8 aborted in state 8. Probably the backend died while processing. client 0 receiving Client 0 aborted in state 10. Probably the backend died while processing. client 7 receiving Client 7 aborted in state 11. Probably the backend died while processing. client 9 receiving Client 9 aborted in state 11. Probably the backend died while processing. client 3 receiving Client 3 aborted in state 10. Probably the backend died while processing. transaction type: TPC-B (sort of) scaling factor: 1 number of clients: 10 number of transactions per client: 1000 number of transactions actually processed: 2025/10000 tps = 15.156125 (including connections establishing) tps = 15.167925 (excluding connections establishing) Is this an expected behaviour? I thought one should have been able to reload the configuration of one of the nodes at any time (for instance, needed when doing an online recovery with PITR as you need to change the archive_command). Shall you need any more information, Tatsuo, please let me know and I'll gladly provide it. -- Jaume Sabater http://linuxsilo.net/ "Ubi sapientas ibi libertas" _______________________________________________ Pgpool-general mailing list [email protected] http://pgfoundry.org/mailman/listinfo/pgpool-general
