I really like SoftLayer. Despite a curious power issue in one part of their data center, their inclusion of KVM over IP on all of their xeon systems is a life jacket that no dedicated server user should be without. I saved a client a full restore thanks to this handy technology.
About 2:30 AM EDT, a box dropped dead. The system has hanging on boot awaiting a disk check for a drive label that could not be found. At many facilities this would have certainly meant a restore. Fortunately, SoftLayer's self-management tools came to the rescue.
Nearly all of SoftLayer's dedicated servers have KVM over IP. KVM over IP allows us to work on the server as if I were standing in front of it with a keyboard and monitor. This can be very powerful in situations where your server does not reboot properly. You first have to use VPN to connect to SL's private network and download Supermicro's IPMI management tools, but once you are logged in, you can now contact the very robust IPMI management card. With it, you have power control, system stats and many other options.
Initially, I thought there may have been a SELinux problem because the system was hanging at:
SELinux: Unregistering netfilter hooks
But as it turned out we simply needed to remove "console=tty0 console=ttyS1,19200n8" from the kernel options due to how the IPMI card redirects the video. Once I removed these options, the full startup scrolled through the screen.
The system hung wanting a disk check and then said it could not find /. Odd thing was / was mounted but did not show up in a "df" output. I checked the /etc/fstab and discovered that the entry for / had gone missing. I know Plesk had some issues with adding quotas and wonder if this may not have been a left over issue from the install and only now surfaced because the server required a fsck.
The key point here is not how to fix this problem but the fact that we could fix it thanks to the robust tools included with most SoftLayer servers.
Had this same issue happened at a provider without KVM over IP solutions, a full server restore would have been required.
To me, this is a very strong selling point. I will certainly be happy when KVM over IP, remote PDUs, virtual media and PXEBoot rescue environments are standard on all systems. Until then, always give SL a look when considering a server.
