Saturday, May 16
Eaten By Mice
In theory, we're scheduled to move to a new server tomorrow.
That might still happen, though it's looking doubtful. The reason it might still happen is that if it doesn't, I'll be up for hundreds of dollars in fees to keep the old servers around for another month.
And the reason it's doubtful is that we had trouble with the systems at work every day last week. Every day, US time. So between 3AM and 5AM my time. And every time, I got woken up and had to fix it.
One hardware fault, one software bug, one software bug in an error-handling routine triggered by another site going haywire, one software bug in normal processing triggered by another site going haywire, and one instance of things just not working right for no apparent reason.
Then I finally got back to setting up the new server. I found some issues with my configuration of the containers for the CPanel sites - insufficent space allocated for the kernel structures, insufficient datagram buffers (which is hardly critical, but I fixed it while I was there) and a couple of other things were we were close to the limit.
So I fixed all of that, recreated the containers, and reinstalled CPanel.
Whereupon I came unstuck, because CPanel would not install.
Turned out - after considerable cursing and deleting and recreating of containers - to be a firewall issue. CPanel's installer couldn't access CPAN (no relation) because CPAN couldn't access the server.
Fucking FTP. Passive mode is there for a reason - though it seems to work about as consistently as secondary DNS servers.
Anyway, sorted that out, but noticed that 6GB of RAM had gone walkies without leaving a forwarding address. It was physically there, but unaccounted for inside Linux. A reboot sorted that out, but then I had a slight accident with a command run on the hardware node instead of inside a container.
So I decided to do a clean reinstall of Linux and load everything again. Since I've been experimenting on the box for two weeks now, this is not overall a bad idea.
The reinstall of CentOS went smoothly. The I went to load OpenVZ - and the OpenVZ repository, and indeed the site - and the list of mirrors - was down.
Gah.
It's back now, but I've pretty much lost a day.
Plan now is to move mee.nu (and the mu.nu Minx sites), and prepare to move mu.nu. That way I can cancel at least one of the servers.
Comments are disabled.
Post is locked.
In theory, we're scheduled to move to a new server tomorrow.
That might still happen, though it's looking doubtful. The reason it might still happen is that if it doesn't, I'll be up for hundreds of dollars in fees to keep the old servers around for another month.
And the reason it's doubtful is that we had trouble with the systems at work every day last week. Every day, US time. So between 3AM and 5AM my time. And every time, I got woken up and had to fix it.
One hardware fault, one software bug, one software bug in an error-handling routine triggered by another site going haywire, one software bug in normal processing triggered by another site going haywire, and one instance of things just not working right for no apparent reason.
Then I finally got back to setting up the new server. I found some issues with my configuration of the containers for the CPanel sites - insufficent space allocated for the kernel structures, insufficient datagram buffers (which is hardly critical, but I fixed it while I was there) and a couple of other things were we were close to the limit.
So I fixed all of that, recreated the containers, and reinstalled CPanel.
Whereupon I came unstuck, because CPanel would not install.
Turned out - after considerable cursing and deleting and recreating of containers - to be a firewall issue. CPanel's installer couldn't access CPAN (no relation) because CPAN couldn't access the server.
Fucking FTP. Passive mode is there for a reason - though it seems to work about as consistently as secondary DNS servers.
Anyway, sorted that out, but noticed that 6GB of RAM had gone walkies without leaving a forwarding address. It was physically there, but unaccounted for inside Linux. A reboot sorted that out, but then I had a slight accident with a command run on the hardware node instead of inside a container.
So I decided to do a clean reinstall of Linux and load everything again. Since I've been experimenting on the box for two weeks now, this is not overall a bad idea.
The reinstall of CentOS went smoothly. The I went to load OpenVZ - and the OpenVZ repository, and indeed the site - and the list of mirrors - was down.
Gah.
It's back now, but I've pretty much lost a day.
Plan now is to move mee.nu (and the mu.nu Minx sites), and prepare to move mu.nu. That way I can cancel at least one of the servers.
Posted by: Pixy Misa at
09:24 PM
| Comments (3)
| Add Comment
| Trackbacks (Suck)
Post contains 424 words, total size 3 kb.
1
Not sure if I approve the new favicon, it's kinda unreadable. But it's distinctive.
Posted by: Pete Zaitcev at Sunday, May 17 2009 01:55 AM (/ppBw)
2
I need to crop it from the larger version; as you say, it's kind of unreadable at that size.
Hang on, I'll do that while I wait for the server to reboot.
Hang on, I'll do that while I wait for the server to reboot.
Posted by: Pixy Misa at Sunday, May 17 2009 02:35 AM (PiXy!)
3
Hrm. That didn't make much difference.
It's supposed to look like this:
So all you need is an 800dpi monitor and a new pair of high-resolution retinas, and you're all set!
It's supposed to look like this:
So all you need is an 800dpi monitor and a new pair of high-resolution retinas, and you're all set!
Posted by: Pixy Misa at Sunday, May 17 2009 02:46 AM (PiXy!)
47kb generated in CPU 0.0157, elapsed 0.1096 seconds.
56 queries taking 0.0988 seconds, 347 records returned.
Powered by Minx 1.1.6c-pink.
56 queries taking 0.0988 seconds, 347 records returned.
Powered by Minx 1.1.6c-pink.