How to configure pool to go down if multiple members are down

Hello pablo_juarez​ 

This is only achievable via external monitor
you can use a script like this

#!/bin/bash
POOL="Your_Pool_name"
THRESHOLD=5

UP_COUNT=$(tmsh show ltm pool $POOL members | grep "Available Members" | awk '{print $4}')

echo "[$(date)] Pool $POOL has $UP_COUNT members up" >> /var/log/pool_monitor.log

if [ "$UP_COUNT" -lt "$THRESHOLD" ]; then
    echo "[$(date)] Pool DOWN: Only $UP_COUNT members up" >> /var/log/pool_monitor.log
    exit 1
else
    echo "[$(date)] Pool UP: $UP_COUNT members up" >> /var/log/pool_monitor.log
    echo "Pool is healthy" 
    exit 0
fi

I was working on this external monitor:

#!/bin/bash
#
# External monitor genérico para marcar DOWN un pool si tiene menos de N miembros UP
#

# Variables de entorno pasadas desde el monitor
POOL=${POOL}
MIN_UP=${MIN_UP:-1}   # por defecto al menos 1

# Obtenemos número de miembros activos del pool
UP=$(tmsh show ltm pool $POOL | grep "Current Active Members" | awk '{print $5}')

# Debug opcional en /var/log/ltm
logger -p local0.info "Monitor $POOL: $UP miembros UP, mínimo requerido $MIN_UP"

if [ "$UP" -lt "$MIN_UP" ]; then
  # No mandamos nada a stdout -> el monitor marcará DOWN
  exit 1
else
  # Mandamos algo (ej: "UP") -> el monitor marcará UP
  echo "UP"
  exit 0
fi

Is it necessary to use at least 2 monitors?
How can the monitor check itself?

I would think that another monitor is needed to verify is members are up or down.

Hi pablo_juarez​,

you can also achieve this with pool member specific monitors. Here is an example with 4 pool members, if 2 or more members go down, the whole pool will go down.

Monitors config, each monitor has a custom destination - its own IP and port.

ltm monitor http mon_http_10_100_153_120 {
    adaptive disabled
    defaults-from http
    destination 10.100.153.120:http
    interval 5
    ip-dscp 0
    recv none
    recv-disable none
    send "GET /\r\n"
    time-until-up 0
    timeout 16
}
ltm monitor http mon_http_10_100_153_121 {
    adaptive disabled
    defaults-from http
    destination 10.100.153.121:http
    interval 5
    ip-dscp 0
    recv none
    recv-disable none
    send "GET /\r\n"
    time-until-up 0
    timeout 16
}
ltm monitor http mon_http_10_100_153_122 {
    adaptive disabled
    defaults-from http
    destination 10.100.153.122:http
    interval 5
    ip-dscp 0
    recv none
    recv-disable none
    send "GET /\r\n"
    time-until-up 0
    timeout 16
}
ltm monitor http mon_http_10_100_153_123 {
    adaptive disabled
    defaults-from http
    destination 10.100.153.123:http
    interval 5
    ip-dscp 0
    recv none
    recv-disable none
    send "GET /\r\n"
    time-until-up 0
    timeout 16
}

 And here goes the pool config, each member has 4 monitors assigned. This way, if two members go down, all pool members go down.

ltm pool pl_custom_monitors {
    members {
        10.100.153.120:http {
            address 10.100.153.120
            monitor min 3 of { mon_http_10_100_153_120 mon_http_10_100_153_121 mon_http_10_100_153_122 mon_http_10_100_153_123 }
            session monitor-enabled
            state up
        }
        10.100.153.121:http {
            address 10.100.153.121
            monitor min 3 of { mon_http_10_100_153_120 mon_http_10_100_153_121 mon_http_10_100_153_122 mon_http_10_100_153_123 }
            session monitor-enabled
            state up
        }
        10.100.153.122:http {
            address 10.100.153.122
            monitor min 3 of { mon_http_10_100_153_120 mon_http_10_100_153_121 mon_http_10_100_153_122 mon_http_10_100_153_123 }
            session monitor-enabled
            state up
        }
        10.100.153.123:http {
            address 10.100.153.123
            monitor min 3 of { mon_http_10_100_153_120 mon_http_10_100_153_121 mon_http_10_100_153_122 mon_http_10_100_153_123 }
            session monitor-enabled
            state up
        }
    }
}

Cheers

Daniel

I havent tested these 2 options that dont need irules, but they might work.

Option1:

Use pool member priority group with Activation less than X.

Assign a high priority value, e.g. 99, to real pool members.
Add dummy dead pool member and assign any lower prio value, e.g. 22

Expectation: When less than X pool members up, vserver will switch to the dummy dead member
but because it's dead then vserver status will be dead too.

Also try without dummy dead.

Option2 (it seems will work):

Use pool's Availability Requirement = At least X,

Create health monitor for each pool member
and specify intended pool member ip addr and port in each of those health monitor's Alias address & Alias port.
Then assign those health monitors into the pool.