Category: Virtualization



Middle of the night on sunday, i get an alert on my phone telling me that the brand new EVA 6300 that i had purchased was experiencing a power supply failure. You can imagine that my heart sank. Shocked about this i logged into my command view VM that monitors my EVA and uses phone home to send it’s alerts to HP and myself.

To my surprise my EVA was telling me that all was clear.  All systems including the power supplies had passed diagnostics. Within a few minutes HP had responded to the error stating that this false positive is a well known issue with the 6300 and will be fixed in the next firmware release.

So for anyone out there experiencing this in the middle of the night… don’t fret it is all ok. If your command view is saying the power supply is ok, but you are receiving  messages stating otherwise, you are probably in the clear.

Advertisements

I was unable to find online information about this, so i thought i would offer up information about this problem in hopes that maybe it will help someone else. During a build update of VMWare along with HBA upgrade and some other hardware enhancements i began to receive random errors from VMWare on one of my hosts regarding an error with the CPU Scheduler. This error racked my brain because there was no loss of functionality to the environment, but i was unwilling to just ignore the random error as some had suggested. the error seemed to stop for a short period of time after a reboot, however nothing ultimately fixed it like a good reinstall does. So as you can imagine where i am going with this, i ended up re installing VMWare on the host, then reapplying the build upgrades and then baam! all has been good since.

So if there is any one out there that is suffering the same fate, fair warning you will need to re install VMWare on the host. This process is only about 20 minutes in time and is actually very easy. Just make sure ALL of your VM’s are off the host because you will want to do a fresh install not a install over the top. Installing over the top of an already corrupt install sometimes turns out worse then expected, so i recommend just start from scratch.


VMWare and other virtual environments are very good at abstracting your hardware and being aware of other hosts in your cluster. One problem i have encountered is when you have an application like ColdFusion running in a VM you have to be careful with what you do to the VM. ColdFusion is a very powerful application, however it is not VMWare or other Virtual Technology aware. It knows the resources assigned to it and the setting provided to it through the JVM. ColdFusion relies on the OS to take care of making sure it’s network paths and hardware are functioning properly. So when you are working in a clustered environment, i have found that CF does not like disruption in it’s application. Even though it recovers just fine from these, sometimes the onset of the initial errors is overwhelming enough that we need a solution to this to sleep better. Now i don’t know all the answers, so i am certainly open to suggestions of what all of you that have encountered the same problem, have found to work. By all means drop a comment and let me know what you have found that works. So here it is.

Once your VM is moved to another host while it is maintaining datasources the links are broken, causing CF to throw a bunch of errors. Not to worry because CF will reestablish it’s DB connection, but if you need to prevent these errors then my recommendation is that either you purchase or setup a load balancer so that you can shift your workload to another server while you move the one VM or you disable maintain datasource connections. Adobe does not recommend you use this method because it creates a lot of overhead for your servers to create and authenticate against a database every time it has to run a query. So by drain stopping the server on the balancer you can remove the users from it so they won’t experience any errors, however if CF is maintaining datasource connections it may still throw an error due to connections being open but no one on them scenario.

%d bloggers like this: