View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0000756 | Pgpool-II | Bug | public | 2022-06-15 18:11 | 2022-06-16 16:38 |
| Reporter | andrei | Assigned To | t-ishii | ||
| Priority | normal | Severity | minor | Reproducibility | always |
| Status | feedback | Resolution | open | ||
| OS | Linux | OS Version | Debian bookworm | ||
| Product Version | 4.3.2 | ||||
| Summary | 0000756: invalid length of startup packet | ||||
| Description | I have setup 2 postgresql-servers (v13.4) with (native) streaming replication and added pgPool2 with the intention of doing load balancing and providing high availability. Everything worked out in regard to load balancing, but after rebooting the primary, it won't get marked with status 'up' when doing "SHOW POOL_NODES;": SHOW POOL_NODES; node_id | hostname | port | status | pg_status | lb_weight | role | pg_role | select_cnt | load_balance_node | replication_delay | replication_state | replication_sync_state | last_status_change ---------+------------+------+--------+-----------+-----------+---------+---------+------------+-------------------+-------------------+-------------------+------------------------+--------------------- 0 | xxxxxxxxxx.2 | 5432 | down | up | 0.000000 | standby | primary | 0 | false | 0 | | | 2022-06-14 14:50:41 1 | xxxxxxxxxx.3 | 5432 | up | up | 1.000000 | standby | standby | 0 | true | 0 | | | 2022-06-14 14:50:41 /etc/pgpool/pgpool.conf backend_hostname0 = 'xxxxxxxx.2' backend_port0 = 5432 backend_weight0 = 0 backend_data_directory0 = '/var/lib/pgsql/data' backend_flag0 = 'ALLOW_TO_FAILOVER' backend_application_name0 = 'db01' backend_hostname1 = 'xxxxxxxx.3' backend_port1 = 5432 backend_weight1 = 2 backend_data_directory1 = '/var/lib/pgsql/data' backend_flag1 = 'ALLOW_TO_FAILOVER' backend_application_name1 = 'db02' When starting pgPool with the -d flag I'm getting this output: 2022-06-15 11:07:49.438: main pid 46280: LOG: find_primary_node_repeatedly: waiting for finding a primary node 2022-06-15 11:07:49.438: child pid 46313: DEBUG: initializing backend status 2022-06-15 11:07:49.438: child pid 46314: DEBUG: initializing backend status 2022-06-15 11:07:49.441: main pid 46280: DEBUG: authenticate kind = 0 2022-06-15 11:07:49.441: main pid 46280: DEBUG: authenticate backend: key data received 2022-06-15 11:07:49.441: main pid 46280: DEBUG: authenticate backend: transaction state: I 2022-06-15 11:07:49.441: main pid 46280: DEBUG: pool_acquire_follow_primary_lock: lock was not held by anyone 2022-06-15 11:07:49.441: main pid 46280: DEBUG: pool_acquire_follow_primary_lock: succeeded in acquiring lock 2022-06-15 11:07:49.441: main pid 46280: DEBUG: do_query: extended:0 query:"SELECT pg_is_in_recovery()" 2022-06-15 11:07:49.442: main pid 46280: DEBUG: do_query: extended:0 query:"SELECT current_setting('server_version_num')" 2022-06-15 11:07:49.443: main pid 46280: DEBUG: get_server_version: backend 1 server version: 130005 2022-06-15 11:07:49.443: main pid 46280: DEBUG: verify_backend_node_status: there's no primary node 2022-06-15 11:07:49.443: main pid 46280: DEBUG: pool_release_follow_primary_lock called 2022-06-15 11:07:49.443: main pid 46280: LOG: find_primary_node: standby node is 1 and postgresql primary reports: 2022-06-14 15:04:35.761 CEST [1050] [unknown]@[unknown] LOG: invalid length of startup packet It's my first bug report so please let me know if I'm missing something. | ||||
| Tags | No tags attached. | ||||
|
|
> Everything worked out in regard to load balancing, but after rebooting the primary, it won't get marked with status 'up' when doing "SHOW POOL_NODES;": I assume that you do not start pgpool with "-D" option. In this case it is possible pgpool found "pool_status" file (usually located under "logdir" of pgpool.conf directive) and the file memorizes that node 0 is in down status (probably while rebooting the server). The file contains something like: down up which means node 0 is in down status. You can: 1) use pcp_attach_node command to "attach" node 0. This should bring node 0 back to online again. 2) edit the file to change "down" to "up" and restart pgpool. Probably you want to choose 0000001 because you don't need to restart pgpool. > and postgresql primary reports: > 2022-06-14 15:04:35.761 CEST [1050] [unknown]@[unknown] LOG: invalid length of startup packet I think this is irrelevant to the issue above. The time stamp is quite bit different. |
