What are you going to do?
What I always do - stay out of trouble... Badly.
Friday, October 28
I popped for the new servers last week, when they had a special offer to waive the $99.95 setup fee.
I got them, confirmed set up properly, today.
Today they have a special offer to double the memory for free. That would save me $500 in the first year. (I decided to only go with 1GB of memory per server because of the upgrade price. If it turns out we need 2GB, that's how much difference it will make.)
I've sent off an email asking if they're willing to do anything for me. Apple were a bunch of doody-heads, but hosting companies rely on continued business, so maybe...
Bumped to the top: THEY SAID YES!!!
Yay! 2x2GB = Munuden!
Update: Memory installed! All systems go! DedicatedNOW rocks my world!
Now I just have to make the application-type stuff work.
Just doing a spot of benchmarking.
Mew is the current server. Comparing it with Kei. The Lovely Angels are equivalent in brainpower* so it doesn't matter which one I test.
|Compress MySQL Backup||25m15.65s||13m9.937s||91.8%|
|Uncompress MySQL Backup||4m27.88s||2m41.47s||65.7%|
|Compress Trackback Log**||53.87s||26.15s||106%|
|Uncompress Trackback Log**||3.65s||2.62s||39.3%|
|Python Loop Test**||5.9s|
The problem with this is hyperthreading. Hyperthreading splits each CPU in half, but Linux doesn't know about this, so just how reflective of reality these results are is somewhat up in the air. The best approach is to run the test many times and pick the lowest number. Or to shut down every other application... Which the Munuvians may not appreciate.
* In this incarnation. Management makes no representations, etc, etc.
** Best of ten trials.
*** Note to self: RPM distributions tend not to be well-optimised. For anything you'll be using a lot - particularly languages - compile your own. It's just a ./configure; make; make install anyway.
Thursday, October 27
I got rather annoyed last night when I received an email informing me that the new servers were configured with 64-bit Linux when this was plainly not true.
It turns out that although they offer 64-bit Linux, you won't get it if you ask for CPanel as well becase there are known problems with that combination (which doesn't surprise me - even though CPanel list it as supported). What they do instead is install a 32-bit kernel specifically optimised for 64-bit processors, which is what they meant when they told me it was 64-bit.
That means that the new servers are ready... For me to start working on.
The billing system is down.
The billing system is down because the billing database is down.
The billing database is down because the database server will not restart.
The database server will not restart because it thinks the shared memory is in use.
It thinks the shared memory is in use because it has a bug related to releasing shared memory segments.
It needed to release the shared memory because it crashed.
It crashed because a user exited while holding a microtransaction lock.
The user exited while holding a microtransaction lock because an index key on another, federated database could not be deleted.
The key could not be deleted because the index is corrupt.
The index is corrupt because the other database crashed.*
The other database crashed because it ran out of disk space.
It ran out of disk space because the database directory was full of backup files.
The directory was full of backup files because if the backup filesystem is not mounted, the backup script writes its files into the source directory.**
The backup filesystem was not mounted because both disks in the RAID set had failed.
The disks in the RAID set had failed because they timed out during reads.
They timed out during reads because there's a bug in the firmware on WD200JB disks.
I don't know why there's a bug in the firmware, perhaps we'll die.
* Even so, this shouldn't happen.
** That one is my fault.
mod_rewrite is your friend. It is not that confusing (once you've spent an hour or two reading the documentation and the examples... twice) and it is very powerful.
For example, if your software only lets you set up blogs as mu.nu/blogname and you really want it to appear at blogname.mu.nu, mod_rewrite will do that for you. You can stick it in an htaccess file, and it will still work (it does this by magic). It works even if blogname.mu.nu is on a different server to mu.nu/blogname. (And if you happen to have a cluster, this means you can effectively use the bandwidth of all of the servers.) You can redirect this but not that if that's what you need. You can even, with a little tweaking, automate the creation of the rulesets.
This may mean that I can use an existing CMS rather than having to write one myself. Further research is definitely indicated.
Wednesday, October 26
The nice people at our new hosting provider have sorted out the clustering and reinstalled the operating system on our new servers.
Only they seem to have reinstalled the 32-bit version.
So it will be another day before I can get started.
On the plus side, we get another freebie: XController, which I was planning to buy. It makes CPanel much simpler and less cluttered. Try the online demo... While we wait for the OS to be reinstalled.
The two new MuNuServers, Kei and Yuri, are almost here. There was a slight mix-up with the operating system (the 32-bit version of CentOS was installed instead of the 64-bit version), and the clustering is waiting on parts... Well, a crossover cable, to be exact.
One bonus: They ran out of 160GB drives, so we got a couple of 200GB drives instead.
It will take a couple of weeks for me to configure and test everything, and then we'll move everything across. This should result in a shinier, happier, more responsive munu for everyone.
Wednesday, October 19
The Daily WTF. It's sort of a funny version of comp.risks:
The Dexia Bank ATM machines are expiriencing a curious problem. The machines stop functioning when someone enters the number 7, making it impossible for people with a 7 in their pin (personal identification number) code to perform a cash withdrawal.I think I know the programmer.
The problem has been occuring for a month. To prevent people from running out of cash, they are able to perform cash withdrawals inside. "We are expiriencing a problem with the software", a Dexia spokesman admitted last wednesday in the daily journal Het Laatste Nieuws, "the problems should be solved within three weeks."
Tuesday, October 18
There's a bug in the firmware of the Western Digital 200JB (also the 200BB, 180JB and 180BB) such that if you use them with a 3Ware 7000 or 8000 series RAID controller, after a certain amount of time (days or weeks) they will cause a spurious timeout and the controller will consider them to have failed.
Mirroring these drives merely gives you a longer window before they disappear.
Well, at least now I know why half a million files just disappeared.
There's a patch. A patch. For a hard disk. Oh, joy.
Sunday, October 16
Weird. I was going to cleam up my blogroll, and I started clicking on some of the links to check them, and a number of them came up with error 412:
Precondition FailedI suspect an overzealous referrer-spam filter, but if that's the case, it's awfully common. Try clicking on any of these links:
The precondition on the request for the URL / evaluated to false.
Oh. Duh. Ambien. I've been cut off from the net. [Sobs.] If I bring up my blog as ai.mu.nu it all works.
58 queries taking 0.2575 seconds, 321 records returned.
Powered by Minx 1.1.6c-pink.