bn8/check_pg_streaming_replication
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

Improve master node output message

Browse source
This commit is contained in:
Benjamin Renard 2020-11-04 15:04:26 +01:00 committed by root
parent dc07eccd09
commit 50c3363cbc
1 changed files with 40 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

53
check_pg_streaming_replication
View file

@ -210,6 +210,20 @@ REPLAY_WARNING_DELAY = $REPLAY_WARNING_DELAY
REPLAY_CRITICAL_DELAY = $REPLAY_CRITICAL_DELAY
"
# Set some stuff to PostgreSQL version
if [ $( echo "$PG_VERSION < 10" |bc -l ) -eq 1 ]
then
pg_current_wal_lsn='pg_current_xlog_location()'
pg_wal_lsn_diff='pg_xlog_location_diff'
sent_lsn='sent_location'
write_lsn='write_location'
else
pg_current_wal_lsn='pg_current_wal_lsn()'
pg_wal_lsn_diff='pg_wal_lsn_diff'
sent_lsn='sent_lsn'
write_lsn='write_lsn'
fi
# Postgres is running ?
if [ $DEBUG -eq 0 ]
then
@ -295,16 +309,16 @@ then
debug "Master application name : $M_APP_NAME"
fi
# Get current state/sync_state from master
M_CUR_STATE_SYNC_STATE="$( psql_master_get "SELECT state,sync_state FROM pg_stat_replication WHERE application_name='$M_APP_NAME';" )"
if [ ! -n "$M_CUR_STATE_SYNC_STATE" ]
# Get current state information from master
M_CUR_REPL_STATE_INFO="$( psql_master_get "SELECT state, sync_state FROM pg_stat_replication WHERE application_name='$M_APP_NAME';" )"
if [ ! -n "$M_CUR_REPL_STATE_INFO" ]
then
echo "UNKNOWN : Can't retreive current state and sync state from master server"
echo "UNKNOWN : Can't retreive current replication state information from master server"
exit 3
fi
debug "Master current state / sync_state : $M_CUR_STATE_SYNC_STATE"
debug "Master current replication state:\n\tstate|sync_state\n\t$M_CUR_REPL_STATE_INFO"
M_CUR_STATE=$( echo "$M_CUR_STATE_SYNC_STATE"|cut -d'|' -f1 )
M_CUR_STATE=$( echo "$M_CUR_REPL_STATE_INFO"|cut -d'|' -f1 )
debug "Master current state : $M_CUR_STATE"
if [ "$M_CUR_STATE" != "streaming" ]
then
@ -312,7 +326,7 @@ then
exit 2
fi
M_CUR_SYNC_STATE=$( echo "$M_CUR_STATE_SYNC_STATE"|cut -d'|' -f2 )
M_CUR_SYNC_STATE=$( echo "$M_CUR_REPL_STATE_INFO"|cut -d'|' -f2 )
debug "Master current sync state : $M_CUR_SYNC_STATE"
if [ "$M_CUR_SYNC_STATE" != "sync" ]
then
@ -375,13 +389,21 @@ else
debug "Postgres is not in recovery mode"
# Check standby client
STANDBY_CLIENTS=$( psql_get "SELECT client_addr, sync_state FROM pg_stat_replication;" )
STANDBY_CLIENTS=$( psql_get "SELECT application_name, client_addr, sent_location, write_location, state, sync_state, current_lag
FROM (
SELECT application_name, client_addr, sent_location, write_location, state, sync_state, current_lag
FROM (
SELECT application_name, client_addr, $sent_lsn AS sent_location, $write_lsn AS write_location, state, sync_state,
$pg_wal_lsn_diff($pg_current_wal_lsn, $write_lsn) AS current_lag
FROM pg_stat_replication
) AS s2
) AS s1" )
if [ ! -n "$STANDBY_CLIENTS" ]
then
echo "WARNING : no stand-by client connected"
exit 1
fi
debug "Stand-by client(s) : $( echo -n $STANDBY_CLIENTS|sed 's/\n/ , /g' )"
debug "Stand-by client(s):\n\t$( echo -e "$STANDBY_CLIENTS"|sed 's/\n/\n\t/' )"
STANDBY_CLIENTS_TXT=""
STANDBY_CLIENTS_COUNT=0
@ -389,11 +411,16 @@ else
do
let STANDBY_CLIENTS_COUNT=STANDBY_CLIENTS_COUNT+1
IP=$( echo $line|cut -d '|' -f 1 )
MODE=$( echo $line|cut -d '|' -f 2 )
STANDBY_CLIENTS_TXT="$STANDBY_CLIENTS_TXT $IP (mode=$MODE)"
NAME=$( echo $line|cut -d '|' -f 1 )
IP=$( echo $line|cut -d '|' -f 2 )
SENT_LOCATION=$( echo $line|cut -d '|' -f 3 )
WRITE_LOCATION=$( echo $line|cut -d '|' -f 4 )
STATE=$( echo $line|cut -d '|' -f 5 )
SYNC_STATE=$( echo $line|cut -d '|' -f 6 )
LAG=$( echo $line|cut -d '|' -f 7 )
STANDBY_CLIENTS_TXT="$STANDBY_CLIENTS_TXT\n$NAME ($IP): $STATE/$SYNC_STATE (Location: sent='$SENT_LOCATION' / write='$WRITE_LOCATION', Lag: ${LAG}b)"
done
echo "OK : $STANDBY_CLIENTS_COUNT stand-by client(s) connected - $STANDBY_CLIENTS_TXT"
echo -e "OK : $STANDBY_CLIENTS_COUNT stand-by client(s) connected\n$STANDBY_CLIENTS_TXT"
exit 0
fi