We have finally released GrokThis Manager 2.0 for our Advanced and Traditional hosting customers! It has been a very long time coming, and in fact it has been ready for some time. The significant change is that it is now web-based. The console version ('gt_manager') remains active and functional.
With our upcoming IP address changes, we are having a number of customers asking about how to keep their site functional when the old IPs drop off.
The simple answer: DNS TTL.
Before the upgrade, simply lower the TTL on your DNS records to something low (say 15 minutes, or 900 seconds). After the upgrade, restore the TTL to its previous setting (say, 28800 or 8 hours). Doing this, users should be directed to the wrong IP for no more than 15 minutes.
We will attempt to implement this for records on our nameservers on the behalf of customers.
While we have already patched our DNS servers for the widely discussed vulnerability; we are concerned that we may have a number of customers that have not patched their servers.
We do advise that all customers update to the latest available version of Bind9, Djbdns, or whatever other DNS server that they may be running. This can normally be done via yum or apt; When doing so, please make sure that the security repositories are enabled in your package manager.
Please be aware that due to a recent SSH vulnerability in Debian & Ubuntu, we have replaced all of our SSH host keys, which may prompt a warning/error when logging into our systems.
We highly advise that customers using a Debian or Ubuntu desktop system create new SSH keys. VPS customers update their OS and replace their OpenSSH host keys. We are in the process of scanning our customer's accounts for insecure keys and will be removing and/or replacing these shortly.
While we understand these actions can cause an inconvenience for many customers, we would rather deal with such inconveniences than with compromised/hacked accounts. We hope that you can understand.
Recently, some users of vps008 have noticed that they cannot complete a reinstall/reimage of Debian or Ubuntu. We resolved the initial complaints by a reboot. Unfortunately, on a VPS server with many customers on it, it can take over an hour to complete a reboot cycle. We hoped that this would not be a recurring issue, but could not be certain of it.
Now, we are noticing that this behavior is, in fact, recurring. Shortly after a reboot, for the following days or weeks, the server will operate normally, but will eventually begin to hang running the configuration for console-setup during a debootstrap installation.
At this point, most customers will eventually close their SSH sessions, and try again to reimage their VPS. When reimaging again, they will get an error that their filesystem is already mounted. This happens because the previous installation never completed.
When our support is contacted, we can stop the prior installation, but without a (time-consuming, service affecting) reboot we cannot provide much further immediate assistance. We can then schedule to reboot the server at a more appropriate time when the service interruption would be less significant to the many other customers that are sharing that host.
We are currently investigating alternative solutions for running the reimages that would not trigger, or be affected by, this bug. Additionally, we are attempting to obtain some assistance from the Xen and Linux developers.
This is a quick notice that we are asking that customers hold off from upgrading GCC to version 4.3 at this time. A problem exists in all current Linux releases (from all vendors) which renders GCC 4.3.0 and higher incompatible. A kernel update will be provided shortly to resolve this issue.
Today, we reached a milestone.... we reached the virtual IRQ limit on VPS008. Unfortunately, this will require that we schedule maintenace in order to issue a kernel upgrade, as the IRQ limit is a hardcoded value in the kernel. I'm sorry to all affected customers, especially considering the recent security-related kernel upgrades made only a couple weeks ago. Customer kernels will NOT be upgraded again at this point, it should hopefully be a quick reboot of our host machine, and back to business!
GrokThis.net customers will be soon receiving emails with login information to our new online billing system, based on Ubersmith. You will be able to login to the billing website, add business contacts, update your credit card, view invoices and payment history, etc...
Unfortunately, this will be YET ANOTHER username and password for you to store. Sorry, you're just gonna have to have another user/pass combo. However, we do hope that you enjoy the new features and new-fangled web management.
Hopefully, we will be able to get better web-based hosting-account management tools online soon. GT_Manager 2.0 is already written, we just gotta get it deployed...
We've been updating our webpages to become more "2.0". The GrokThis.net webpage now has keyword-based navigation, and the VPS Village pages are entirely AJAX-enabled.
Quite simply, we're using Scriptalicious and Prototype for most of it. There is very little new backend code. Its quite simply a matter of removing the templates from the code. Out backend layout something like:
The content div will now update automatically, and it will fail-over for older browsers! Mason's idea of components really works well with the AJAX concept and there are all kinds of things you can do with it, and in much prettier ways than this. This example is meant to be short and easy... and very important, simple for retrofitting an existing site. Have fun, and take care!
One of our customers has posted an interesting article on his blog, Making the most out of a 64MB VPS. He explains a few of the difficulties that can be encountered in a limited memory environment, shows how to decrease your memory usage, view performance statistics, and more...
My biggest suggestion, personally, however... if you're going for an account with only 64 or 96MB of ram, you will see performance advantages from going with a 32-bit OS. The only reason we don't make it default, is that a mix-arch Xen installation was sort of experimental until very recently.
We reviewed whether or not ntpd was updating time on servers and found 3 machines where it was not updating correctly and restarted it. We are making plans to configure a local ntp pool.
We've updated the security settings on the non-default SMTP port 26 on Gimli to be restricted only to authenticated access, and allowing for a larger message size. This is how we have most/all of our servers configured, this change makes Gimli behave as our other servers do.
Customers may use port 26 for SMTP to bypass restrictive ISP firewalls and to send larger-sized messages. The security setting prevents email from arriving on this port, allowing only customers to connect.
In addition to standard IM based support, we now have a 'live support' link on the webpage that will initiate a live instant messanger online, without the need for an IM account!
We've updated the skin for community.grokthis.net and made it easier to subscribe to RSS feeds for various things. For instance, when you are viewing a topic/tag (click the topic/tab), you can then click to subscribe to that topic.