Note: I first wrote this a month or two ago, so it is possible a newer version of HCIBench has fixed some of these issues.
This is definitely not an exhaustive troubleshooting guide HCIBench for networking….however I did want to provide a couple of tips that might save you some time…..they definitely would have helped me.
The first of which to be aware of is to just be aware of this button in the HCIBench gui.
If this box isn’t checked, all of the “helper” VMs that HCIBench spins up will try to get their IP via DHCP. This itself isn’t that crazy as the tooltip that appears for this option pretty much states that. The thing to be aware of is how often you are running HCIBench or even doing a validation before a test. Each time HCIBench is doing a validation or a test, it is pulling an IP from your DHCP.
Just after a couple of runs, I had 12 new ip leases in DHCP. If one is not careful it can quickly fill up your DHCP pool(I wasn’t careful). Then someone would need to figure out why suddenly HCIBench was no longer able to grab IPs…..
Now the second thing and issue I ran into is:
Huh, looks familiar.
With the option selected, HCI bench becomes its own DHCP server and statically assigns the Ips to the “helper” vms as it sees fit.
I am sure for a decent number of people this is not an issue, but it can be if your HCIBench is configured like this:
From above, you see that the DNS server is on a 192.168.1 subnet and the regular address of the HCIBench server is on a 10.10.1 subnet. Nothing too crazy right?
Well I found that when I did this, during Validation HCIBench would lose all connectivity and fail Validation…and be unavailable for several minutes afterward.
Really annoying, well after some troubleshooting and some digging I found a comment on the HCIBench website that is a workaround to this very issue. I will let Mick Kelleher’s comment do the explaining.
Long story short, depending on how you set your ip config for HCIBench, if you also have the option checked from above, HCIBench may freak out and not know what to do.
Fortunately Mick also provided the workaround above.
After utilizing the work around, we see the helper VMs get an IP on a completely different subnet and my Validation and test was able to complete successfully.
Hopefully this helps someone out there 🙂