I am presently dealing with a problem that I had not anticipated. I have deployed SME server multiple times and have always been pleased with its utter simplicity and ease of installation. But this ease and simplicity comes at a cost of being unable to prevent tragic mistakes as they occur.
In my case, I had not realized that when a system is being installed, it will enlist USB drives into RAID building if they are available without asking the user or otherwise showing a summary of changes about to be committed. It does, however, ask if you would like to recover from backup, including USB, leading one to believe that USB drives are identified at install time as possible sources of data to be restored. This was my assumption about the install process... and no, I have no recall of reading otherwise in the manual and I am pretty sure I read the manual from cover to cover. While I know my memory is faulty, I simply don't recall any warning to remove any USB drives not intended for use in a RAID system creation.
But regardless of this, every Linux distro I have ever experienced gives the user a review of the partitioning layout prior to committing the change. SME server does not. For people experienced in the use and deployment of SME server, this is no big deal as the behavior is already understood. But for people who are not, this is a potentially fatal problem that could be avoided with a summary screen stating "the following is the partition layout... [partition layout table] Is this okay?" Seeing this, I would have realized the mistake I was about to make and halted immediately. Instead, the drive was repartitioned and enlisted as a 3rd drive in a mirrored RAID configuration.
But what lead me to this terrible position? Well, it was an RPM dependency problem that should be resolved at the CentOS level. I wanted to remove the Japanese language extensions from the installation. I used the web interface. The dependencies were such that removing the Japanese language elements pulled virtually the entire e-smith system out of the installation immediately and without any prompting -- it just went and that was all. I was able to ssh into the box, but recovery was impossible in a short time. So I decided to rely on backing up the data or on previously completed backups.
I attempted to "upgrade" but it did nearly nothing as the system seemed rather unaffected after the very quick upgrade process ran. I ran an install and... well... the first paragraph details what I am experiencing.
The USB drive's partitions were changed from a single ext3 partition to two RAID partitions as are observable in a typical SME installation. The file root directory structure, however, seems to remain showing file and directory names... those files and directories are, however, unavailable and the OS reports an I/O error due to unreachable sectors. I suspect repartitioning will enable some level of recovery as I don't believe the drive was set up as anything but a spare in the RAID mirror... I hope that is the case anyway.
This lengthy account is a plea to include some sort of summary table describing the partition layout including which drives are being enlisted into the configuration prior to committing the partitioning. It could save someone like me... and I suspect there are many people at least as stupid wanting to use SME server. The reputation of Linux is built on experiences like these and if I were not a very experienced Linux user, I would blame Linux as a very risky and unforgiving system to use. My experience with Ubuntu, Fedora and RHEL/CentOS has set my expectations and I see that SME Server is very different in this regard and I don't think it is a good idea.
10 Replies
1937 Views