Forum
Welcome, Guest
Username: Password: Remember me

TOPIC:

Ha-lizard on the XCP-NG 8.2 in 2021. Progress of my deployment 2 weeks 1 day ago #2455

I see great resistance from the XCP community whenever "ha-lizard" is mentioned, especially when referring to "splitbrain". They always try to promote XCP's own native features. But this is the magic of free software. There are infinite scenarios and HA-LIZARD is the best that has adapted to my case.

Please Log in or Create an account to join the conversation.

Ha-lizard on the XCP-NG 8.2 in 2021. Progress of my deployment 1 week 6 days ago #2456

I haven't seen any split brain events other than during my redundancy and stress tests years ago when I purposely triggered it.

More than happy to help push it... I've been running a low cost production environment for several years now using community based products like ISPconfig and HA-Lizard has been extremely valuable for peace of mind. I really welcome the dual low cost server approach rather that super expensive HA clusters.

I think there is a lot of fear about split brain scenario and said misconceptions regarding 2 node DRBD. Maybe a good place to start is to alleviate these misconceptions via XCP and XO community forums. I'm actually quite surprised that the developers havne't taken an interest given the practical nature of the HA-Lizard project. Surely it would be an asset to them.

Please Log in or Create an account to join the conversation.

Ha-lizard on the XCP-NG 8.2 in 2021. Progress of my deployment 1 week 1 hour ago #2461

h!!
I have on my server nic0 and nic1 for the exclusive use of VMs. Nic2 and nic3 in bond for DRBD / iSCSI and
NIC4 for management. During the initial replication process (watch cat / proc / drbd command) I removed each cabling and none affected the replication process. Replication just stopped when I removed the management cable (nic4). The question is: shouldn't replication occur by nic2 and nic3?

Please Log in or Create an account to join the conversation.

Ha-lizard on the XCP-NG 8.2 in 2021. Progress of my deployment 4 days 21 hours ago #2462

  • Salvatore Costantino
  • Salvatore Costantino's Avatar
  • Offline
  • Posts: 680
Any chance you set the management interface IPs as the replication IPs?

Also, check that you selected the correct interface during installation. Try the below and make sure that the configured interface name for "DRBD_INTERFACE" has been configured correctly. It should be set to the name of the replication link.
cat /etc/iscsi-ha/iscsi-ha.conf

Please Log in or Create an account to join the conversation.

Ha-lizard on the XCP-NG 8.2 in 2021. Progress of my deployment 3 days 2 hours ago #2465

I checked using the command you showed me and the interface is correct. It is xapi3 and with the ifconfig command it is possible to see that it is the IP configured in the replication bond.
I will do more tests on this.

Now a new question / problem. I had two VMs, and HA enabled, but when removing the power from the master (stress test), should the VMs have migrated to the slave? I did this test after the initial sync. Should I have activated or verified any more services?

Please Log in or Create an account to join the conversation.

Ha-lizard on the XCP-NG 8.2 in 2021. Progress of my deployment 3 days 1 hour ago #2466

  • Salvatore Costantino
  • Salvatore Costantino's Avatar
  • Offline
  • Posts: 680
Yes, the VMs should have been started on the surviving host, however, a number of conditions must me met in order for that to happen.

- HA (halizard HA and not XCP/Xenserver HA) must be enabled. If you enabled HA within XCP, then halizard would automatically disable itself
- your heuristic IP address must be reachable during the event

The ha-lizard logs will provide full details of what occurred on the surviving host. Can you check /var/log/user.log around the time of the test and post the log lines here.

Please Log in or Create an account to join the conversation.