From: Hitoshi H. <hem...@la...> - 2012-07-11 07:01:26
|
Hi list, We have been testing GTM HA feature, and found a problem: once gtm got failed, if someone or some AP accessed to coordinator before gtm failover complete, gtm_standby can never be reconnect to the cluster. ============================================================================ POSTGRES-XC BUG REPORT TEMPLATE ============================================================================ Your name : Hitoshi Hemmi Your email address : hem...@la... System Configuration: --------------------- Architecture : x86_64 Operating Systems : CentOS release 6.2 x86_64 Postgres-XC version : Postgres-XC 1.0 Compilers used : gcc (GCC) 4.4.6 20110731 (Red Hat 4.4.6-3) Description of problems: ============================================== 1. Simmulated failure of gtm gtm_ctl stop -Z gtm -m immediate -D /home/postgres/data/gtm 2. Unluckily someone access to a coordinator psql -p 5432 -c "SELECT * FROM pgxc_node;" 3. Promotion of gtm_standby gtm_ctl promote -Z gtm -D /home/postgres/data/standby 4. Try to reconnet ... gtm_ctl reconnect -Z gtm_proxy -D /home/postgres/data/gtm_proxy -o '-s localhost -t 7777' 5. BANG! WARNING: can not connect to GTM: Connection refused FATAL: Can not register Datanode on GTM WARNING: can not connect to GTM: Connection refused WARNING: can not connect to GTM: Connection refused WARNING: Xid is invalid. WARNING: can not connect to GTM: Connection refused WARNING: can not connect to GTM: Connection refused ERROR: GTM error, could not obtain snapshot ... -- Hitoshi HEMMI NTT Open Source Software Center hem...@la... (Please note that my address has changed.) Tel:(03)5860-5115 Fax:(03)5463-5490 |