View Issue Details

IDProjectCategoryView StatusLast Update
0000465Pgpool-IIBugpublic2019-03-05 04:11
ReportercaudatusAssigned Topengbo 
PrioritynormalSeveritymajorReproducibilityalways
Status assignedResolutionopen 
PlatformLinuxOSCentOSOS Version7
Product Version4.0.2 
Target VersionFixed in Version 
Summary0000465: pgpool cannot accept new connections while attached backend stream-wal replicated dbslave:tcp:5432 unreachable
Description#################################################################################
os: CentOS7 x86_64
    
pgpool: 3.7.7
    4.0.2

struct: pgpool + dbmaster (pg95) + stream-wal replication + dbslave (pg95)
node states:
    dbmaster: up/connected
    dbslave: up/connected

config:
    health_check_period = 0
    sr_check_period = 0
    load_balance_mode = off
    master_slave_mode = on
    master_slave_sub_mode = 'stream'
    replication_mode = off

event:
    if attached node: dbslave's network goes down immediately (iptables ... -j DROP)

incident:
    then pgpool:
        cannot accept new psql connections (it seems as an hangup in accepting new connections) (BAD BEHAVIOUR!)
        while serves all queries continuesly in all existing connections (GOOD.)

usecase:
    backend struct: master-slave pg backends + direct stream-wal postgres replication
    with config:
        load-balance: on or off
        master_slave_sub_mode: stream
        health_check: enabled or disabled
        sr_check: enabled or disabled
        
    we need pgpool must:
        serve:
            - read-write queries (via primary master)
            - and accept new connections continueusly and serves queries (via primary master)
            
        independently:
            - if slave is in any state (up, down, waiting, etc...)
            - if slave:tcp:5432 reachable or unreachable

#################################################################################

note:
    if slave postgresql was shutdown (pg_ctl stop command), behaviour is right:
        - pgpool serves existing connections and accept new connections
        - and serves them continueusly
        
    bad behaviour happens only when attached slave node's network goes down (aka pgpool cannot reach slave:tcp:5432 through network)
Steps To Reproduce- start pgpool, dbmaster, dbslave
- backend_nodes are up
- # iptables -A net-net -p tcp -m tcp -s $GPGOOLIP -d $DBSLAVEIP --dport 5432 -j DROP
- # while true; timeout 3 psql -U $PGUSER -h $PGPOOLHOST postgres -c 'select pg_is_in_recovery()' || echo TIMEDOUT; sleep 1; done

it gives TIMEDOUT, until unreachable dbslave node will be detached. (BAD)
existing psql connections served continueusly (GOOD)
TagsNo tags attached.

Activities

caudatus

2019-02-15 02:46

reporter  

bugpack.tgz (120,999 bytes)

caudatus

2019-02-15 03:10

reporter   ~0002386

maybe this issue descripts the same bug as 0000177, but with extended informations

caudatus

2019-02-15 22:07

reporter   ~0002387

one more note for this report:
- if i switch off load_balance_mode, healt_check and sr_check (i think, it should mean, that there's nothing to communicate on client queries for pgpool to dbslave node),
  there is even an empty connection (authorized) request and immediately a connection close request from pgpool to dbslave (it )

psql client query:
root@pgproxy4-dbtest-plus-dp1:/etc/pgpool-II# export PGPASSWORD=xxx; while true; do { echo -en "$(date):\t" ; timeout 5 psql -A -t -P pager=off -h -U pgpoolcheck postgres -c '\l' || echo TIMEDOUT; } | head -1; sleep 1; done
Fri Feb 15 12:52:09 UTC 2019: postgres|postgres|SQL_ASCII|C|C|=Tc/postgres
...

csv log on dbslave node:
2019-02-15 12:52:09.782 UTC,,,7599,"pgproxy-dbtest.docker1:48218",5c66b5f9.1daf,1,"",2019-02-15 12:52:09 UTC,,0,LOG,00000,"connection received: host=pgproxy-dbtest.docker1 port=48218",,,,,,,,,""
2019-02-15 12:52:09.805 UTC,"pgpoolcheck","postgres",7599,"pgproxy-dbtest.docker1:48218",5c66b5f9.1daf,2,"authentication",2019-02-15 12:52:09 UTC,2/47747,0,LOG,00000,"connection authorized: user=pgpoolcheck database=postgres",,,,,,,,,""
2019-02-15 12:52:09.810 UTC,"pgpoolcheck","postgres",7599,"pgproxy-dbtest.docker1:48218",5c66b5f9.1daf,3,"idle",2019-02-15 12:52:09 UTC,,0,LOG,00000,"disconnection: session time: 0:00:00.029 user=pgpoolcheck database=postgres host=pgproxy-dbtest.docker1 port=48218",,,,,,,,,"psql"

I don't know why is this connection attempt?
I think this connect-disconnect-to-dbslave operation can be the real reason of hangup.

However, if the psql test queries contains one persistent connect attempt and then sqeuencial queries as:
root@pgproxy4-dbtest-plus-dp1:/etc/pgpool-II# export PGPASSWORD=xxx; while true; do psql -A -t -h localhost -U pgpoolcheck postgres < <(while true; do echo 'select now(), pg_is_in_recovery();'; sleep 1; done;); sleep 1; done
2019-02-15 13:02:55.659116+00|f
2019-02-15 13:02:56.632916+00|f
...

Also there is a connection attempt on dbslave, but (certainly) there's no any queries:
csv.log:
2019-02-15 13:02:55.640 UTC,,,7614,"pgproxy-dbtest.docker1:48520",5c66b87f.1dbe,1,"",2019-02-15 13:02:55 UTC,,0,LOG,00000,"connection received: host=pgproxy-dbtest.docker1 port=48520",,,,,,,,,""
2019-02-15 13:02:55.656 UTC,"pgpoolcheck","postgres",7614,"pgproxy-dbtest.docker1:48520",5c66b87f.1dbe,2,"authentication",2019-02-15 13:02:55 UTC,2/47756,0,LOG,00000,"connection authorized: user=pgpoolcheck database=postgres",,,,,,,,,""

and if meanwhile pgpool lost dbslave:tc:5432, pgpool can serve queries without any forced operation to dbslave

caudatus

2019-02-15 23:13

reporter   ~0002388

- if i reduce connect_timeout from 10000 to 10, even the bug bad behaviour is the same :(

pengbo

2019-02-19 15:06

developer   ~0002393

Could you show your pgpool.conf?
I would like to confirm the config of "health_check_timeout".

caudatus

2019-02-19 20:52

reporter   ~0002394

yes, it's in attached bugpack.tgz

pgpool.conf:
# ----------------------------
# pgPool-II configuration file
# ----------------------------
#
# This file consists of lines of the form:
#
# name = value
#
# Whitespace may be used. Comments are introduced with "#" anywhere on a line.
# The complete list of parameter names and allowed values can be found in the
# pgPool-II documentation.
#
# This file is read on server startup and when the server receives a SIGHUP
# signal. If you edit the file on a running system, you have to SIGHUP the
# server for the changes to take effect, or use "pgpool reload". Some
# parameters, which are marked below, require a server shutdown and restart to
# take effect.
#


#------------------------------------------------------------------------------
# CONNECTIONS
#------------------------------------------------------------------------------

# - pgpool Connection Settings -

listen_addresses = '*'
port = 5432
#socket_dir = '/var/run/postgresql'
socket_dir = '/tmp'
                                   # Unix domain socket path
                                   # The Debian package defaults to
                                   # /var/run/postgresql
                                   # (change requires restart)
listen_backlog_multiplier = 2
                                   # Set the backlog parameter of listen(2) to
                                   # num_init_children * listen_backlog_multiplier.
                                   # (change requires restart)
serialize_accept = off
                                   # whether to serialize accept() call to avoid thundering herd problem
                                   # (change requires restart)

# - pgpool Communication Manager Connection Settings -

pcp_listen_addresses = '*'
                                   # Host name or IP address for pcp process to listen on:
                                   # '*' for all, '' for no TCP/IP connections
                                   # (change requires restart)
pcp_port = 9898
                                   # Port number for pcp
                                   # (change requires restart)
#pcp_socket_dir = '/var/run/postgresql'
pcp_socket_dir = '/tmp'
                                   # Unix domain socket path for pcp
                                   # The Debian package defaults to
                                   # /var/run/postgresql
                                   # (change requires restart)

# - Backend Connection Settings -

#backend_hostname1 = 'host2'
#backend_port1 = 5433
#backend_weight1 = 1
#backend_data_directory1 = '/data1'
#backend_flag1 = 'ALLOW_TO_FAILOVER'

# - Authentication -

enable_pool_hba = on
pool_passwd = 'pool_passwd'
                                   # File name of pool_passwd for md5 authentication.
                                   # "" disables pool_passwd.
                                   # (change requires restart)
authentication_timeout = 60
                                   # Delay in seconds to complete client authentication
                                   # 0 means no timeout.

allow_clear_text_frontend_auth = off
                                   # Allow Pgpool-II to use clear text password authentication
                                   # with clients, when pool_passwd does not
                                   # contain the user password


# - SSL Connections -

ssl = on


#------------------------------------------------------------------------------
# POOLS
#------------------------------------------------------------------------------

# - Concurrent session and pool size -

num_init_children = 1
                                   # Number of concurrent sessions allowed
                                   # (change requires restart)
max_pool = 4
                                   # Number of connection pool caches per connection
                                   # (change requires restart)

# - Life time -

child_life_time = 300
                                   # Pool exits after being idle for this many seconds
child_max_connections = 0
                                   # Pool exits after receiving that many connections
                                   # 0 means no exit
connection_life_time = 0
                                   # Connection to backend closes after being idle for this many seconds
                                   # 0 means no close
client_idle_limit = 0
                                   # Client is disconnected after being idle for that many seconds
                                   # (even inside an explicit transactions!)
                                   # 0 means no disconnection


#------------------------------------------------------------------------------
# LOGS
#------------------------------------------------------------------------------

# - Where to log -

log_destination = 'syslog'

# - What to log -

log_line_prefix = '%t: pid %p: ' # printf-style string to output at beginning of each log line.

log_connections = on
log_hostname = on
log_statement = on
log_per_node_statement = on
log_standby_delay = 'always'
                                   # Log standby delay
                                   # Valid values are combinations of always,
                                   # if_over_threshold, none

# - Syslog specific -

syslog_facility = 'LOCAL0'
                                   # Syslog local facility. Default to LOCAL0
syslog_ident = 'pgpool'
                                   # Syslog program identification string
                                   # Default to 'pgpool'

# - Debug -

#log_error_verbosity = default # terse, default, or verbose messages
log_error_verbosity = verbose

client_min_messages = warning # values in order of decreasing detail:
                                        # debug5
                                        # debug4
                                        # debug3
                                        # debug2
                                        # debug1
                                        # log
                                        # notice
                                        # warning
                                        # error

log_min_messages = info # values in order of decreasing detail:
                                        # debug5
                                        # debug4
                                        # debug3
                                        # debug2
                                        # debug1
                                        # info
                                        # notice
                                        # warning
                                        # error
                                        # log
                                        # fatal
                                        # panic

#------------------------------------------------------------------------------
# FILE LOCATIONS
#------------------------------------------------------------------------------

pid_file_name = '/var/run/pgpool/pgpool.pid'
                                   # PID file name
                                   # Can be specified as relative to the"
                                   # location of pgpool.conf file or
                                   # as an absolute path
                                   # (change requires restart)
logdir = '/var/log/pgpool'
                                   # Directory of pgPool status file
                                   # (change requires restart)


#------------------------------------------------------------------------------
# CONNECTION POOLING
#------------------------------------------------------------------------------

connection_cache = off
                                   # Activate connection pools
                                   # (change requires restart)

                                   # Semicolon separated list of queries
                                   # to be issued at the end of a session
                                   # The default is for 8.3 and later
reset_query_list = 'ABORT; DISCARD ALL'
                                   # The following one is for 8.2 and before
#reset_query_list = 'ABORT; RESET ALL; SET SESSION AUTHORIZATION DEFAULT'


#------------------------------------------------------------------------------
# REPLICATION MODE
#------------------------------------------------------------------------------

replication_mode = off
                                   # Activate replication mode
                                   # (change requires restart)
replicate_select = off
                                   # Replicate SELECT statements
                                   # when in replication mode
                                   # replicate_select is higher priority than
                                   # load_balance_mode.

insert_lock = off
                                   # Automatically locks a dummy row or a table
                                   # with INSERT statements to keep SERIAL data
                                   # consistency
                                   # Without SERIAL, no lock will be issued
lobj_lock_table = ''
                                   # When rewriting lo_creat command in
                                   # replication mode, specify table name to
                                   # lock

# - Degenerate handling -

replication_stop_on_mismatch = off
                                   # On disagreement with the packet kind
                                   # sent from backend, degenerate the node
                                   # which is most likely "minority"
                                   # If off, just force to exit this session

failover_if_affected_tuples_mismatch = off
                                   # On disagreement with the number of affected
                                   # tuples in UPDATE/DELETE queries, then
                                   # degenerate the node which is most likely
                                   # "minority".
                                   # If off, just abort the transaction to
                                   # keep the consistency


#------------------------------------------------------------------------------
# LOAD BALANCING MODE
#------------------------------------------------------------------------------

load_balance_mode = off
ignore_leading_white_space = on
                                   # Ignore leading white spaces of each query
white_function_list = ''
                                   # Comma separated list of function names
                                   # that don't write to database
                                   # Regexp are accepted
black_function_list = 'nextval,setval,nextval,setval,insert,update,alter'

black_query_pattern_list = ''
                                   # Semicolon separated list of query patterns
                                   # that should be sent to primary node
                                   # Regexp are accepted
                                   # valid for streaming replicaton mode only.

database_redirect_preference_list = ''
                                   # comma separated list of pairs of database and node id.
                                   # example: postgres:primary,mydb[0-4]:1,mydb[5-9]:2'
                                   # valid for streaming replicaton mode only.
app_name_redirect_preference_list = ''
                                   # comma separated list of pairs of app name and node id.
                                   # example: 'psql:primary,myapp[0-4]:1,myapp[5-9]:standby'
                                   # valid for streaming replicaton mode only.
allow_sql_comments = off
                                   # if on, ignore SQL comments when judging if load balance or
                                   # query cache is possible.
                                   # If off, SQL comments effectively prevent the judgment
                                   # (pre 3.4 behavior).

disable_load_balance_on_write = 'transaction' # Load balance behavior when write query is issued
                        # in an explicit transaction.
                        # Note that any query not in an explicit transaction
                        # is not affected by the parameter.
                        # 'transaction' (the default): if a write query is issued,
                        # subsequent read queries will not be load balanced
                        # until the transaction ends.
                        # 'trans_transaction': if a write query is issued,
                        # subsequent read queries in an explicit transaction
                        # will not be load balanced until the session ends.
                        # 'always': if a write query is issued, read queries will
                        # not be load balanced until the session ends.

#------------------------------------------------------------------------------
# MASTER/SLAVE MODE
#------------------------------------------------------------------------------

master_slave_mode = on
master_slave_sub_mode = 'stream'

# - Streaming -

sr_check_period = 0
sr_check_user = 'pgpcheck'

sr_check_password = 'xxx'

sr_check_database = 'postgres'
                                   # Database name for streaming replication check
delay_threshold = 0
                                   # Threshold before not dispatching query to standby node
                                   # Unit is in bytes
                                   # Disabled (0) by default

# - Special commands -

follow_master_command = '/usr/bin/logger -t pgpoolalert "follow_master id:%d.%h.%M, new id:%m.%M"'
                                   # Executes this command after master failover
                                   # Special values:
                                   # %d = node id
                                   # %h = host name
                                   # %p = port number
                                   # %D = database cluster path
                                   # %m = new master node id
                                   # %H = hostname of the new master node
                                   # %M = old master node id
                                   # %P = old primary node id
                                   # %r = new master port number
                                   # %R = new master database cluster path
                                   # %% = '%' character

#------------------------------------------------------------------------------
# HEALTH CHECK GLOBAL PARAMETERS
#------------------------------------------------------------------------------

health_check_period = 0
health_check_timeout = 3
health_check_user = 'pgpcheck'
health_check_password = 'xxx'

health_check_database = 'postgres'
                                   # Database name for health check. If '', tries 'postgres' frist, then 'template1'

health_check_max_retries = 3
                                   # Maximum number of times to retry a failed health check before giving up.
health_check_retry_delay = 1
                                   # Amount of time to wait (in seconds) between retries.
connect_timeout = 1000
                                   # Timeout value in milliseconds before giving up to connect to backend.
                                   # Default is 10000 ms (10 second). Flaky network user may want to increase
                                   # the value. 0 means no timeout.
                                   # Note that this value is not only used for health check,
                                   # but also for ordinary conection to backend.

#------------------------------------------------------------------------------
# HEALTH CHECK PER NODE PARAMETERS (OPTIONAL)
#------------------------------------------------------------------------------
#health_check_period0 = 0
#health_check_timeout0 = 20
#health_check_user0 = 'nobody'
#health_check_password0 = ''
#health_check_database0 = ''
#health_check_max_retries0 = 0
#health_check_retry_delay0 = 1
#connect_timeout0 = 10000

#------------------------------------------------------------------------------
# FAILOVER AND FAILBACK
#------------------------------------------------------------------------------

failover_command = '/usr/bin/logger -t pgpoolalert "failover id:%d.%h.%M, new id:%m.%M"'
#failover_command = '/srv/postgres-data/attach "%d" | logger -t pgpool_failover "%d"'

                                   # Executes this command at failover
                                   # Special values:
                                   # %d = node id
                                   # %h = host name
                                   # %p = port number
                                   # %D = database cluster path
                                   # %m = new master node id
                                   # %H = hostname of the new master node
                                   # %M = old master node id
                                   # %P = old primary node id
                                   # %r = new master port number
                                   # %R = new master database cluster path
                                   # %% = '%' character
failback_command = '/usr/bin/logger -t pgpoolalert "failback id:%d.%h.%M, new id:%m.%M"'
                                   # Executes this command at failback.
                                   # Special values:
                                   # %d = node id
                                   # %h = host name
                                   # %p = port number
                                   # %D = database cluster path
                                   # %m = new master node id
                                   # %H = hostname of the new master node
                                   # %M = old master node id
                                   # %P = old primary node id
                                   # %r = new master port number
                                   # %R = new master database cluster path
                                   # %% = '%' character

failover_on_backend_error = off

detach_false_primary = off

#search_primary_node_timeout = 300
search_primary_node_timeout = 0
                                   # Timeout in seconds to search for the
                                   # primary node when a failover occurs.
                                   # 0 means no timeout, keep searching
                                   # for a primary node forever.

#------------------------------------------------------------------------------
# ONLINE RECOVERY
#------------------------------------------------------------------------------

recovery_user = 'nobody'
                                   # Online recovery user
recovery_password = ''
                                   # Online recovery password
                                   # Leaving it empty will make Pgpool-II to first look for the
                                   # Password in pool_passwd file before using the empty password

recovery_1st_stage_command = ''
                                   # Executes a command in first stage
recovery_2nd_stage_command = ''
                                   # Executes a command in second stage
recovery_timeout = 90
                                   # Timeout in seconds to wait for the
                                   # recovering node's postmaster to start up
                                   # 0 means no wait
client_idle_limit_in_recovery = 0
                                   # Client is disconnected after being idle
                                   # for that many seconds in the second stage
                                   # of online recovery
                                   # 0 means no disconnection
                                   # -1 means immediate disconnection


#------------------------------------------------------------------------------
# WATCHDOG
#------------------------------------------------------------------------------

# - Enabling -

use_watchdog = off
                                    # Activates watchdog
                                    # (change requires restart)

# -Connection to up stream servers -

trusted_servers = ''
                                    # trusted server list which are used
                                    # to confirm network connection
                                    # (hostA,hostB,hostC,...)
                                    # (change requires restart)
ping_path = '/bin'
                                    # ping command path
                                    # (change requires restart)

# - Watchdog communication Settings -

wd_hostname = ''
                                    # Host name or IP address of this watchdog
                                    # (change requires restart)
wd_port = 9000
                                    # port number for watchdog service
                                    # (change requires restart)
wd_priority = 1
                                    # priority of this watchdog in leader election
                                    # (change requires restart)

wd_authkey = ''
                                    # Authentication key for watchdog communication
                                    # (change requires restart)

#wd_ipc_socket_dir = '/var/run/postgresql'
wd_ipc_socket_dir = '/tmp'
                                    # Unix domain socket path for watchdog IPC socket
                                    # The Debian package defaults to
                                    # /var/run/postgresql
                                    # (change requires restart)


# - Virtual IP control Setting -

delegate_IP = ''
                                    # delegate IP address
                                    # If this is empty, virtual IP never bring up.
                                    # (change requires restart)
if_cmd_path = '/sbin'
                                    # path to the directory where if_up/down_cmd exists
                                    # (change requires restart)
if_up_cmd = 'ip addr add $_IP_$/24 dev eth0 label eth0:0'
                                    # startup delegate IP command
                                    # (change requires restart)
if_down_cmd = 'ip addr del $_IP_$/24 dev eth0'
                                    # shutdown delegate IP command
                                    # (change requires restart)
arping_path = '/usr/sbin'
                                    # arping command path
                                    # (change requires restart)
arping_cmd = 'arping -U $_IP_$ -w 1'
                                    # arping command
                                    # (change requires restart)

# - Behaivor on escalation Setting -

clear_memqcache_on_escalation = on
                                    # Clear all the query cache on shared memory
                                    # when standby pgpool escalate to active pgpool
                                    # (= virtual IP holder).
                                    # This should be off if client connects to pgpool
                                    # not using virtual IP.
                                    # (change requires restart)
wd_escalation_command = ''
                                    # Executes this command at escalation on new active pgpool.
                                    # (change requires restart)
wd_de_escalation_command = ''
                                    # Executes this command when master pgpool resigns from being master.
                                    # (change requires restart)

# - Watchdog consensus settings for failover -

failover_when_quorum_exists = off

failover_require_consensus = on
                                    # Perform failover when majority of Pgpool-II nodes
                                    # aggrees on the backend node status change
                                    # (change requires restart)

allow_multiple_failover_requests_from_node = off
                                    # A Pgpool-II node can cast multiple votes
                                    # for building the consensus on failover
                                    # (change requires restart)

# - Lifecheck Setting -

# -- common --

wd_monitoring_interfaces_list = '' # Comma separated list of interfaces names to monitor.
                                    # if any interface from the list is active the watchdog will
                                    # consider the network is fine
                                    # 'any' to enable monitoring on all interfaces except loopback
                                    # '' to disable monitoring
                                    # (change requires restart)


wd_lifecheck_method = 'heartbeat'
                                    # Method of watchdog lifecheck ('heartbeat' or 'query' or 'external')
                                    # (change requires restart)
wd_interval = 10
                                    # lifecheck interval (sec) > 0
                                    # (change requires restart)

# -- heartbeat mode --

wd_heartbeat_port = 9694
                                    # Port number for receiving heartbeat signal
                                    # (change requires restart)
wd_heartbeat_keepalive = 2
                                    # Interval time of sending heartbeat signal (sec)
                                    # (change requires restart)
wd_heartbeat_deadtime = 30
                                    # Deadtime interval for heartbeat signal (sec)
                                    # (change requires restart)
heartbeat_destination0 = 'host0_ip1'
                                    # Host name or IP address of destination 0
                                    # for sending heartbeat signal.
                                    # (change requires restart)
heartbeat_destination_port0 = 9694
                                    # Port number of destination 0 for sending
                                    # heartbeat signal. Usually this is the
                                    # same as wd_heartbeat_port.
                                    # (change requires restart)
heartbeat_device0 = ''
                                    # Name of NIC device (such like 'eth0')
                                    # used for sending/receiving heartbeat
                                    # signal to/from destination 0.
                                    # This works only when this is not empty
                                    # and pgpool has root privilege.
                                    # (change requires restart)

#heartbeat_destination1 = 'host0_ip2'
#heartbeat_destination_port1 = 9694
#heartbeat_device1 = ''

# -- query mode --

wd_life_point = 3
                                    # lifecheck retry times
                                    # (change requires restart)
wd_lifecheck_query = 'SELECT 1'
                                    # lifecheck query to pgpool from watchdog
                                    # (change requires restart)
wd_lifecheck_dbname = 'template1'
                                    # Database name connected for lifecheck
                                    # (change requires restart)
wd_lifecheck_user = 'nobody'
                                    # watchdog user monitoring pgpools in lifecheck
                                    # (change requires restart)
wd_lifecheck_password = ''
                                    # Password for watchdog user in lifecheck
                                    # Leaving it empty will make Pgpool-II to first look for the
                                    # Password in pool_passwd file before using the empty password
                                    # (change requires restart)

# - Other pgpool Connection Settings -

#other_pgpool_hostname0 = 'host0'
                                    # Host name or IP address to connect to for other pgpool 0
                                    # (change requires restart)
#other_pgpool_port0 = 5432
                                    # Port number for other pgpool 0
                                    # (change requires restart)
#other_wd_port0 = 9000
                                    # Port number for other watchdog 0
                                    # (change requires restart)
#other_pgpool_hostname1 = 'host1'
#other_pgpool_port1 = 5432
#other_wd_port1 = 9000


#------------------------------------------------------------------------------
# OTHERS
#------------------------------------------------------------------------------
relcache_expire = 0
                                   # Life time of relation cache in seconds.
                                   # 0 means no cache expiration(the default).
                                   # The relation cache is used for cache the
                                   # query result against PostgreSQL system
                                   # catalog to obtain various information
                                   # including table structures or if it's a
                                   # temporary table or not. The cache is
                                   # maintained in a pgpool child local memory
                                   # and being kept as long as it survives.
                                   # If someone modify the table by using
                                   # ALTER TABLE or some such, the relcache is
                                   # not consistent anymore.
                                   # For this purpose, cache_expiration
                                   # controls the life time of the cache.

relcache_size = 256
                                   # Number of relation cache
                                   # entry. If you see frequently:
                                   # "pool_search_relcache: cache replacement happend"
                                   # in the pgpool log, you might want to increate this number.

check_temp_table = on
                                   # If on, enable temporary table check in SELECT statements.
                                   # This initiates queries against system catalog of primary/master
                                   # thus increases load of master.
                                   # If you are absolutely sure that your system never uses temporary tables
                                   # and you want to save access to primary/master, you could turn this off.
                                   # Default is on.

check_unlogged_table = on
                                   # If on, enable unlogged table check in SELECT statements.
                                   # This initiates queries against system catalog of primary/master
                                   # thus increases load of master.
                                   # If you are absolutely sure that your system never uses unlogged tables
                                   # and you want to save access to primary/master, you could turn this off.
                                   # Default is on.

#------------------------------------------------------------------------------
# IN MEMORY QUERY MEMORY CACHE
#------------------------------------------------------------------------------
memory_cache_enabled = on
                                   # If on, use the memory cache functionality, off by default
memqcache_method = 'shmem'
                                   # Cache storage method. either 'shmem'(shared memory) or
                                   # 'memcached'. 'shmem' by default
                                   # (change requires restart)
memqcache_memcached_host = 'localhost'
                                   # Memcached host name or IP address. Mandatory if
                                   # memqcache_method = 'memcached'.
                                   # Defaults to localhost.
                                   # (change requires restart)
memqcache_memcached_port = 11211
                                   # Memcached port number. Mondatory if memqcache_method = 'memcached'.
                                   # Defaults to 11211.
                                   # (change requires restart)
memqcache_total_size = 67108864
                                   # Total memory size in bytes for storing memory cache.
                                   # Mandatory if memqcache_method = 'shmem'.
                                   # Defaults to 64MB.
                                   # (change requires restart)
memqcache_max_num_cache = 1000000
                                   # Total number of cache entries. Mandatory
                                   # if memqcache_method = 'shmem'.
                                   # Each cache entry consumes 48 bytes on shared memory.
                                   # Defaults to 1,000,000(45.8MB).
                                   # (change requires restart)
memqcache_expire = 0
                                   # Memory cache entry life time specified in seconds.
                                   # 0 means infinite life time. 0 by default.
                                   # (change requires restart)
memqcache_auto_cache_invalidation = on
                                   # If on, invalidation of query cache is triggered by corresponding
                                   # DDL/DML/DCL(and memqcache_expire). If off, it is only triggered
                                   # by memqcache_expire. on by default.
                                   # (change requires restart)
memqcache_maxcache = 409600
                                   # Maximum SELECT result size in bytes.
                                   # Must be smaller than memqcache_cache_block_size. Defaults to 400KB.
                                   # (change requires restart)
memqcache_cache_block_size = 10485760
                                   # Cache block size in bytes. Mandatory if memqcache_method = 'shmem'.
                                   # Defaults to 1MB.
                                   # (change requires restart)
memqcache_oiddir = '/var/log/pgpool/oiddir'
                                      # Temporary work directory to record table oids
                                   # (change requires restart)
white_memqcache_table_list = ''
                                   # Comma separated list of table names to memcache
                                   # that don't write to database
                                   # Regexp are accepted
black_memqcache_table_list = ''
                                   # Comma separated list of table names not to memcache
                                   # that don't write to database
                                   # Regexp are accepted
ssl_key = '/etc/pgpool-II/server-ssl.key'
ssl_cert = '/etc/pgpool-II/server-ssl.crt'

backend_hostname1 = '172.18.0.12'
backend_port1 = 5432
backend_weight1 = 1
backend_data_directory1 = '/srv/postgres-data/data'
backend_flag1 = 'ALLOW_TO_FAILOVER'

backend_hostname0 = '172.18.0.11'
backend_port0 = 5432
backend_weight0 = 1
backend_data_directory0 = '/srv/postgres-data/data'
backend_flag0 = 'ALLOW_TO_FAILOVER'

parallel_mode = false

caudatus

2019-02-19 20:56

reporter   ~0002395

It doesn't matter if health_check is on or off,
On every new connection request pgpool tries to connect (w. auth) and disconnect onto slave backend also
I tried several variations of configuration options and pgpool showed this behaviour in all cases.

pengbo

2019-02-26 11:29

developer   ~0002406

This is an expected behavior of Pgpool-II,
because non blocking connect() system call is used.

If the network is cut by using "iptables", config of "connect_timeout" doesn't work
and the connection will wait until TCP/IP timeout.

caudatus

2019-03-05 04:11

reporter   ~0002418

Thank You for your response, but it held a sad news for us.

1. Is it any config variation to allow pgpool receive new client connections while network between pgpool -> backend_slave is down in any reason?
We need no load_balance, health_check and replication_check at all and replication process remains directly between backends through stream replication.

2. Why does pgpool try to forward client's new connect and authentication request to backend_slave when load_balance=off?...
It's just an empty `connect w auth + disconnect` sequence without any operation...

Issue History

Date Modified Username Field Change
2019-02-15 02:46 caudatus New Issue
2019-02-15 02:46 caudatus File Added: bugpack.tgz
2019-02-15 03:10 caudatus Note Added: 0002386
2019-02-15 10:20 administrator Assigned To => pengbo
2019-02-15 10:20 administrator Status new => assigned
2019-02-15 22:07 caudatus Note Added: 0002387
2019-02-15 23:13 caudatus Note Added: 0002388
2019-02-19 15:06 pengbo Note Added: 0002393
2019-02-19 20:52 caudatus Note Added: 0002394
2019-02-19 20:56 caudatus Note Added: 0002395
2019-02-26 11:29 pengbo Note Added: 0002406
2019-03-05 04:11 caudatus Note Added: 0002418