TOPIC:

Hi,
I have installed HA-Lizard/HA-ISCSI on two hosts for testing purposes with the nosan installer. Each host has 4 NICs. One for management (10.0.0.101-2), two bonded for DRBD(10.10.10.1-2) and one for the VM (10.0.0.103).

When i poweroff one host the VM on it successfully failovers to the other.

1)When i disconnect the management ethernet, the vm on the primary host starts on the second but it keeps running on the primary. I read in another topic that it needs hardware fencing. Shouldn't just fence the primary after it loses quorum?
2) When i disconnect the ethernet that the vm uses (10.0.0.103) nothing happens.
3) If i disconnect one of the bonded ethernets nothing happens again and it seems it doesn't replicate any data (or it breaks the database of my vm in the second host).

My config:
DISABLED_VAPPS=()
ENABLE_LOGGING=1
FENCE_ACTION=stop
FENCE_ENABLED=1
FENCE_FILE_LOC=/etc/ha-lizard/fence
FENCE_HA_ONFAIL=1
FENCE_HEURISTICS_IPS=10.0.0.171
FENCE_HOST_FORGET=0
FENCE_IPADDRESS=
FENCE_METHOD=POOL
FENCE_MIN_HOSTS=2
FENCE_PASSWD=
FENCE_QUORUM_REQUIRED=1
FENCE_REBOOT_LONE_HOST=0
FENCE_USE_IP_HEURISTICS=1
GLOBAL_VM_HA=1
MAIL_FROM="root@localhost"
MAIL_ON=1
MAIL_SUBJECT="SYSTEM_ALERT-FROM_HOST:$HOSTNAME"
MAIL_TO="root@localhost"
MONITOR_DELAY=15
MONITOR_KILLALL=1
MONITOR_MAX_STARTS=20
MONITOR_SCANRATE=10
OP_MODE=2
PROMOTE_SLAVE=1
SLAVE_HA=1
SLAVE_VM_STAT=1
SMTP_PASS=""
SMTP_PORT="25"
SMTP_SERVER="127.0.0.1"
SMTP_USER=""
XAPI_COUNT=2
XAPI_DELAY=10
XC_FIELD_NAME='ha-lizard-enabled'
XE_TIMEOUT=10

Thank you in advance!

1)When i disconnect the management ethernet, the vm on the primary host starts on the second but it keeps running on the primary. I read in another topic that it needs hardware fencing. Shouldn't just fence the primary after it loses quorum?
This scenario can only be reliably handled by using HW fencing where the slave would first power-off the master before starting any VMs.

2) When i disconnect the ethernet that the vm uses (10.0.0.103) nothing happens.
VM Ethernet ports fall outside of the scope of HA. HA in our case specifically deals with (1) HW failure (2) VM failure.

3) If i disconnect one of the bonded ethernets nothing happens again and it seems it doesn't replicate any data (or it breaks the database of my vm in the second host).
There could be something wrong with your bonding setup. Disconnecting one of the bonded ports should not affect replication at all.

We have a test release prepared that addresses the test case of the management link being pulled from the master. To work, you must have FENCE_REBOOT_LONE_HOST set to 1. Upon loss of the management link the Master will self fence. Assuming the slave is still alive - it will promote itself to Master and run ALL VMs. When the master comes back (management link restored) - it will demote itself to slave.

Please test and report back your results. ** FYI - the proper way to address this scenario is still to use STONITH.

SW can be downloaded here:
halizard.org/release/ha-lizard/ha-lizard-1.8.9.tc01.tgz

Unpack the tarball in /tmp and then run the install script on top of your existing setup. All your settings will be preserved.