Myself, along with two of my colleagues, went on a day trip to this year’s IPEXPO Earls Court in London where we were shown a demo of esXpress at PHD Virtual Technology’s booth. We were very impressed by the presentation and some of the futures of esXpress. We were especially interested in the File Level Recovery features of their data De-Duplication appliance.
When I got back home from IPEXPO on Thursday, I decided to test the product for myself. So, I went on to download a 30 day evaluation from http://www.phdvirtual.com/ and I can surely say that I’ve been putting PHD Virtual esXpress 3.6 through some vigorous testing for the past few days. To start off with, I found esXpress easy to deploy and I was quite impressed with it. However I must confess that when I really started to dig into how esXpress goes about its business I started to have some concerns about it. However, I have decided not make any my concerns public until I have had a good chat with some people at PHD Virtual as I think that they will have an answer to most of my questions.
Now the actual reason for this post is to make a note (mostly for my own reference) regarding a technical issue that I ran into during my esXpress testing. The issue relates to File Level Recovery (FLR) using the esXpress DeDupe appliance. When trying to do a FLR from a backed up VMDK with Reiserfs file systems that were created on VMs running Suse Linux Enterprise Server 10, the DeDupe appliance crashes with a “kernel panic” fatal error. I have flagged this up with PHD Virtual Support and they came back with the following statement:
The commit i/o error on ext3 is working as designed. That is the ext3 filesystem trying to replay the journal and we are not allowing it as this would be modifying your backed up data. The issue with reiserfs appears to be the exact same thing, except that the kernel is crashing rather than handling the error gracefully like ext3 does. Technically, this is a bug in reiserfs. That said, we are working on a solution to resolve this issue.
For updates on the issue, refer to the following thread: