TOPIC:

Afternoon,

I regularly get a notification that the Slave in my 2 host cluster is unreachable via HTTP even though I'm working on it at the time. Not all the time obviously but I do get it periodically. Why do I get this alert? I get this repetitively but I know there will be an update to resolve that.

When we get emails when a VM is shutdown/started up I get the the UUID for the vm. The alert is fine but can I have the VM name from VM Properties General detail? It would just make more sense as to what has gone down?

Cheers
Simon

We check whether XAPI is live on slaves two ways in quick succession: first send a ping to the management port of the slave - if successful then make sure XAPI is responding to HTTP on the management port. If HTTP does not succeed - we enter a loop until it returns OR trigger HA if it does not.

Is it possible XAPI is becoming slow or unresponsive at times on your slave? The above logic would trigger an email with this content:

"Pool Host on Server: <UUID> not responding to HTTP - manual intervention may be required"

Can you verify this is the email you are receiving.. If so, it may be due to XAPI being slow to respond. Is it possible this is occuring during live migrations or some other network load?

Regarding email notices in general - we will be porting new mail handling functionality from our HA-iSCSI project to HA-Lizard. This will prevent repetitive emails from being sent for issues that have already been alerted. This will be in the next patch release - likely 2-4 weeks away.

Your point about sending the VM name in addition to the UUID on email alerts is a good one. We'll look at rolling this into the next release also.

Pulse Supply wrote: Can you verify this is the email you are receiving.. If so, it may be due to XAPI being slow to respond. Is it possible this is occuring during live migrations or some other network load?

Yes that is the alert & No not really. Currently running two VM's with low network requirements. I am running a single vlan into the network port on the host server. I'm letting the Spanning tree auto calculate on this port. I could set it to portfast so it would know to operate as a host client instead of a possible switch or other networking device. I'd just expect this would be the same for the master as they are configured the same. They are in Gb ports so I certainly wouldn't expect throttling at network level.

Anything else I can check in the meantime? XAPI related?

How many CPUs are made available to dom0? We've seen poor performance when only a single CPU is allocated for dom0. If you have enough resources, a default installation of XCP/XenServer should show 4 CPUs.

Simon,
We have duplicated the issue with the HTTP unreachable alert you receive. It is being caused by a an interrupted system call (outside of ha-lizard) which suddently closes a socket used to check if xapi is live on a slave. A fix will be rolled into a patch release once we isolate where the interrupt is originating from.

Any idea on what this means also?

Failed to update global pool configuration settings in /etc/ha-lizard/ha-lizard.pool.conf - Check Configuration!

Thanks again,
Simon