No boom today. Boom tomorrow. There's always a boom tomorrow.
Saturday, July 21
Proves The Rule, I Guess...
If you want comments on your site, post something wrong.
If you want comments on your site, post something wrong.
Anyway, to those who haven't visited before: Welcome, and yes, I know that WWI was basically a family feud among Queen Victoria's grandchildren, and that The Molotov-Ribbentrop Pact would be a good name for a rock band.
Ancient Chinese history I'm less reliable on, so if you want to take up a discussion of the Yellow Turban Rebellion in my comments, go right ahead.
Posted by: Pixy Misa at
12:08 AM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 79 words, total size 1 kb.
Sunday, February 05
Design Refresh
I'm doing a new design for the next version of Minx, based on the 960.gs / Skeleton / Bootstrap CSS layout libraries.*
Update: Damn arithmetic! One problem with the above layout is that to fit ads neatly in the sidebar you'd want it to be 240 pixels wide - the same as the ad itself. But the maths just doesn't work out.
* Most likely Bootstrap; I had some issues with version 1.4, but the newly released 2.0 cleans up most of the things I didn't like and adds even more features.
I'm doing a new design for the next version of Minx, based on the 960.gs / Skeleton / Bootstrap CSS layout libraries.*
The rounded corners are likely to go at this stage; form design will improve, and blogs will resize (at least in theory) to fit your device, but in discrete steps rather than one pixel at a time.
The idea is that you'll choose a 12- or 16-column layout, and then assign a certain number of columns to each element on the page, so you might choose 12 columns, and allocate 8 to the content and 4 to the sidebar. But you could also have a headlines area (between the banner and the content) with three items each four columns wide.
The new base widths will be 700 pixels (for smaller devices like tablets and phones), 940 pixels (for older PCs and notebooks), and 1180 pixels (for larger screens). All of those work out evenly whether you choose a 12- or 16-column grid.
There will be a pair of new, interactive menu bars above and below your banner image, the top one for the mee.nu system as a whole, the bottom one for your site. The current ads (which I haven't sold any of yet anyway) will shrink down to fit in the top menu bar, rather than sitting above it, and will expand out on mouseover. I think that's the best compromise to make them as unobtrusive as possible while still giving advertisers a useful amount of space.
Sample Images
With a 940-pixel standard layout, you have 12 columns each 60 pixels wide, and 11 margins in between each 20 pixels wide. 12 x 60 + 11 x 20 = 940.
With 16 columns, it's 16 x 40 + 15 x 20 = 940.
This works because we're ignoring the rightmost 20-pixel margin - if we included that, the widths would be 720, 960, and 1200 pixels - all multiples of 240, with lots and lots of useful factors.
So if you have a 3 column sidebar in a 12-column layout, that's 3 x 60 + 2 x 20 = 220px. 4 columns in 16-col layout is 4 x 40 + 3 x 20 = 220px. Either way, too narrow for the ad. 4 columns in 12-col layout is 4 x 60 + 3 x 20 = 300px; 5 columns in 16-col layout is 5 x 40 + 3 x 20 = 280px, which leaves a fair chunk of space over.
I'm not sure how bad that will be in a live design, so I'm not going to tear up the fundamental principles of mathematics just yet. And you could force the sidebar into a 240-pixel layout within a 280/300 pixel division if need be, with a larger than normal gap between the sidebar and the main content.
Real-world testing is indicated here.
Posted by: Pixy Misa at
10:04 AM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 547 words, total size 4 kb.
Tuesday, August 30
Location, Location, Location
Okay, I've lined up reasonably-priced providers in the following locations:
One of the providers in Phoenix - I/O Flood - is also a tiny company, but they clearly know what they are doing (very active online) and their pricing is great. The other two providers have been in business for years and I've had servers with both of them previously, but they're the most expensive. (But still significantly cheaper than what I'm paying now.)
Fremont, Dallas, and New Jersey would make sense from a geographical perspective, but I really like the guys in Phoenix. Maybe Phoenix and Atlanta.
Ah, decisions, decisions.
Okay, I've lined up reasonably-priced providers in the following locations:
- Phoenix (2)
- Dallas (2)
- Atlanta
- New Jersey
- Chicago
- Los Angeles
- Fremont
One of the providers in Phoenix - I/O Flood - is also a tiny company, but they clearly know what they are doing (very active online) and their pricing is great. The other two providers have been in business for years and I've had servers with both of them previously, but they're the most expensive. (But still significantly cheaper than what I'm paying now.)
Fremont, Dallas, and New Jersey would make sense from a geographical perspective, but I really like the guys in Phoenix. Maybe Phoenix and Atlanta.
Ah, decisions, decisions.
Posted by: Pixy Misa at
10:08 AM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 155 words, total size 1 kb.
Saturday, August 27
Planned Downtime
We're going to have a little downtime around 1AM CDT this Sunday to replace the failed drive. Should only be around 15 minutes assuming all goes well. If all goes wrong, could be a couple of hours. I'd place the probabilities at about 60/40.
Running and double-checking backups right now.
Update: Ran a little late and a little long and a little hiccupy, but all done now.
Update to the update: And then when all was said and done, the network decided to notwork: The switch lost track of all our IP addresses and effectively null-routed the server. Had to manually bind them directly to the network interface, then restart the virtual servers one by one.
We're going to have a little downtime around 1AM CDT this Sunday to replace the failed drive. Should only be around 15 minutes assuming all goes well. If all goes wrong, could be a couple of hours. I'd place the probabilities at about 60/40.
Running and double-checking backups right now.
Update: Ran a little late and a little long and a little hiccupy, but all done now.
Update to the update: And then when all was said and done, the network decided to notwork: The switch lost track of all our IP addresses and effectively null-routed the server. Had to manually bind them directly to the network interface, then restart the virtual servers one by one.
Posted by: Pixy Misa at
02:09 PM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 118 words, total size 1 kb.
Sunday, August 21
Sorry About That, Folks!
Had a drive failure1, which caused the system to reboot2, which caused a lengthy filesystem check because we'd been up so long3, which when circumvented allowed us to boot cleanly, which highlighted a networking glitch4, which when fixed allowed us to find the database corruption5 which we've now fixed allowing us to get back on the air.
I was supposed to be in a conference session on hosting web applications right about now...
1 Grrr Western Digital! 6
2 Grrr Adaptec! 6
3 Grrr ext3 default settings which I didn't bother to override...6
4 Grrr OpenVZ! 6
5 Grrr MySQL! 6
6 Though it must be said that we've been running on this without a hiccup for nearly 18 months.
Had a drive failure1, which caused the system to reboot2, which caused a lengthy filesystem check because we'd been up so long3, which when circumvented allowed us to boot cleanly, which highlighted a networking glitch4, which when fixed allowed us to find the database corruption5 which we've now fixed allowing us to get back on the air.
I was supposed to be in a conference session on hosting web applications right about now...
1 Grrr Western Digital! 6
2 Grrr Adaptec! 6
3 Grrr ext3 default settings which I didn't bother to override...6
4 Grrr OpenVZ! 6
5 Grrr MySQL! 6
6 Though it must be said that we've been running on this without a hiccup for nearly 18 months.
Posted by: Pixy Misa at
09:26 AM
| Comments (7)
| Add Comment
| Trackbacks (Suck)
Post contains 124 words, total size 1 kb.
Thursday, January 27
(Tinkery-Type Noises)
Youtube video, default size (now widescreen, 480x270 16:9 video).
[youtube=5_sfnQDr1-o]
Youtube video, custom size, including the extra 30 vertical pixels the player uses:
[youtube=5_sfnQDr1-o size=400x255]
Youtube video, custom size, not including the extra 30 vertical pixels - the BBCode engine detects if you've entered something that works out to exactly 4:3 or 16:9 and adjusts the player size appropriately:
[youtube=5_sfnQDr1-o size=400x225]
If you mess up the size it still works:
[youtube=5_sfnQDr1-o size=400x224]
Specify just the width - assumes you want 16:9:
[youtube=5_sfnQDr1-o size=540x]
Youtube video, default size (now widescreen, 480x270 16:9 video).
[youtube=5_sfnQDr1-o]
Youtube video, custom size, including the extra 30 vertical pixels the player uses:
[youtube=5_sfnQDr1-o size=400x255]
Youtube video, custom size, not including the extra 30 vertical pixels - the BBCode engine detects if you've entered something that works out to exactly 4:3 or 16:9 and adjusts the player size appropriately:
[youtube=5_sfnQDr1-o size=400x225]
If you mess up the size it still works:
[youtube=5_sfnQDr1-o size=400x224]
Specify just the width - assumes you want 16:9:
[youtube=5_sfnQDr1-o size=540x]
Posted by: Pixy Misa at
06:38 AM
| Comments (2)
| Add Comment
| Trackbacks (Suck)
Post contains 76 words, total size 1 kb.
Wednesday, June 09
Spring Cleaning
While in fact it's Wintertime here at Pixy Labs (and it's bloody freezing - or close to, anyway), I've just spent a couple of hours dusting off the main mee.nu site and I'm going to be posting there again. In fact, I'm going to try posting there every day. Even if it kills me. Which it probably will.
Going to put up a new banner too. I can recycle the current one next month anyway.
While in fact it's Wintertime here at Pixy Labs (and it's bloody freezing - or close to, anyway), I've just spent a couple of hours dusting off the main mee.nu site and I'm going to be posting there again. In fact, I'm going to try posting there every day. Even if it kills me. Which it probably will.
Going to put up a new banner too. I can recycle the current one next month anyway.
Posted by: Pixy Misa at
12:37 AM
| Comments (1)
| Add Comment
| Trackbacks (Suck)
Post contains 77 words, total size 1 kb.
Friday, April 23
Testing Stuffs
Test test...
more...
Test test...
more...
Posted by: Pixy Misa at
02:36 AM
| Comments (7)
| Add Comment
| Trackbacks (Suck)
Post contains 7 words, total size 1 kb.
Saturday, April 17
Server Migration - Sunday 18th April 8AM-9AM GMT
We'll be moving to our shiny new server this Sunday at 6PM Sydney time - 8AM GMT, 4AM EDT, 1AM PDT. So a mostly convenient time for most people - in the evening for me, and in the middle of the night for almost everyone else.*
Downtime is expected to be 30 to 60 minutes, depending on exactly how long the network switch takes to refresh its ARP cache. All URLs, logins, and IP address will be unchanged; it should just go offline and then come back an hour later on the new server running (up to) twice as fast.
Update: Pushed back one day and bumped. My test migration worked perfectly, but I only got there half an hour before the real thing was scheduled to happen, so I'm putting it off until tomorrow.
* 8 o'clock on a Sunday morning counts as the middle of the night according to ISO 8833.7.
We'll be moving to our shiny new server this Sunday at 6PM Sydney time - 8AM GMT, 4AM EDT, 1AM PDT. So a mostly convenient time for most people - in the evening for me, and in the middle of the night for almost everyone else.*
Downtime is expected to be 30 to 60 minutes, depending on exactly how long the network switch takes to refresh its ARP cache. All URLs, logins, and IP address will be unchanged; it should just go offline and then come back an hour later on the new server running (up to) twice as fast.
Update: Pushed back one day and bumped. My test migration worked perfectly, but I only got there half an hour before the real thing was scheduled to happen, so I'm putting it off until tomorrow.
* 8 o'clock on a Sunday morning counts as the middle of the night according to ISO 8833.7.
Posted by: Pixy Misa at
05:37 PM
| No Comments
| Add Comment
| Trackbacks (Suck)
Post contains 161 words, total size 1 kb.
Friday, July 17
One Last Outage - Around 1AM CDT
Today's outage is proudly brought to you by Adaptec - "The number one name in overpriced RAID controllers with crappy firmware" - and Western Digital - "Our drives are so tough they eat other disk drives! Also, your data."
This is to replace the insane drive and wipe and rebuild the corrupted RAID-5 volume. Should be about 15 minutes. Once it's done, everything should be good - we have a RAID-5 array and a RAID-1 array and an enterprise SSD for the databases. What could possibly go wrong?
...
No, thanks, I don't want a list.
Update: DONE!
Update: And for some bizarre reason, after the maintenance was done and I'd made sure everything was working correctly, the networking went notworking. Bleh.
Today's outage is proudly brought to you by Adaptec - "The number one name in overpriced RAID controllers with crappy firmware" - and Western Digital - "Our drives are so tough they eat other disk drives! Also, your data."
This is to replace the insane drive and wipe and rebuild the corrupted RAID-5 volume. Should be about 15 minutes. Once it's done, everything should be good - we have a RAID-5 array and a RAID-1 array and an enterprise SSD for the databases. What could possibly go wrong?
...
No, thanks, I don't want a list.
Update: DONE!
Update: And for some bizarre reason, after the maintenance was done and I'd made sure everything was working correctly, the networking went notworking. Bleh.
Posted by: Pixy Misa at
02:13 PM
| No Comments
| Add Comment
| Trackbacks (Suck)
Post contains 129 words, total size 1 kb.
Friday, June 19
And We're Back
I said half an hour, and half an hour it took!
Update: Except that my clever trick of redirecting the reverse proxy to minimise downtime caused the spam filter to block, um, everyone.
Most people should have picked up the DNS changes by now and won't have that problem any more.
I said half an hour, and half an hour it took!
Update: Except that my clever trick of redirecting the reverse proxy to minimise downtime caused the spam filter to block, um, everyone.
Most people should have picked up the DNS changes by now and won't have that problem any more.
Posted by: Pixy Misa at
08:30 PM
| Comments (12)
| Add Comment
| Trackbacks (Suck)
Post contains 54 words, total size 1 kb.
Thursday, June 18
Really Moving
The on-again off-again mee.nu server move is on again, for this Friday at 6AM GMT.
The on-again off-again mee.nu server move is on again, for this Friday at 6AM GMT.
Posted by: Pixy Misa at
07:09 PM
| No Comments
| Add Comment
| Trackbacks (Suck)
Post contains 17 words, total size 1 kb.
Monday, June 08
Mooving
mee.nu (and the mu.nu Minx blogs) will be migrating across to its new home on Monday at 6AM GMT.
Was originally scheduled for Sunday at the same time, but I somehow slept through all of Saturday, so I'm not quite ready yet.
Downtime is expected to be about an hour. Originally it was going to be just a few minutes, but due to recent hiccups I'm now planning to do a full database reload, so it will take a little longer.
Update: Running a bit behind schedule, bit still planning to do the move today. I have worked out a plan to minimise the downtime, so when your blog goes blank for twenty minutes or so, that will be it.
(I should be able to put up a message saying what's happening.)
Update: Still working on it. Tidying up the old environment now so that I can do a nice quick move. While I'm doing that, I'll do a test run of the database load.
Update: Downtime should be only about 15 minutes. The trial run took 3.5 minutes to dump the database, 1 minute to copy across to the new server, and 8.5 minutes to load. That's not at all bad.
The setup is nearly ready, so I'm going to proceed and get it done before the database glitches again.
Update: Damn. The database dump doesn't work correctly if there are corrupt indexes, so I'll have to do an index rebuild first, then dump.
Update: Ran into a bug. Looks like it's not happening today after all. I'll keep working on it and reschedule, probably to tomorrow.
Update: Doh! I know what's going on. Hang on...
Update: Nope, that wasn't it.
Weird. I'm already running a CherryPy app in production on this server, so why won't this one start? I can't even get a trivial sample program to run. Same version of Python and CherryPy too. Hmm.
Update: Foo. Networking wasn't running correctly on the newly-created virtual server. I restarted it, and everything is fine. It looked like a firewall issue - with the new system, not only is the server protected from the outside world, but all the virtual servers are protected from each other, and, in fact, from themselves. But no, it was just that the loopback interface wasn't running.
Update: Okay, looks like the application server and database are working fine now. Just need to set up the load-balancer. But it's nearly midnight, so I won't be doing the move today; it will be tomorrow instead.
mee.nu (and the mu.nu Minx blogs) will be migrating across to its new home on Monday at 6AM GMT.
Was originally scheduled for Sunday at the same time, but I somehow slept through all of Saturday, so I'm not quite ready yet.
Downtime is expected to be about an hour. Originally it was going to be just a few minutes, but due to recent hiccups I'm now planning to do a full database reload, so it will take a little longer.
Update: Running a bit behind schedule, bit still planning to do the move today. I have worked out a plan to minimise the downtime, so when your blog goes blank for twenty minutes or so, that will be it.
(I should be able to put up a message saying what's happening.)
Update: Still working on it. Tidying up the old environment now so that I can do a nice quick move. While I'm doing that, I'll do a test run of the database load.
Update: Downtime should be only about 15 minutes. The trial run took 3.5 minutes to dump the database, 1 minute to copy across to the new server, and 8.5 minutes to load. That's not at all bad.
The setup is nearly ready, so I'm going to proceed and get it done before the database glitches again.
Update: Damn. The database dump doesn't work correctly if there are corrupt indexes, so I'll have to do an index rebuild first, then dump.
Update: Ran into a bug. Looks like it's not happening today after all. I'll keep working on it and reschedule, probably to tomorrow.
Update: Doh! I know what's going on. Hang on...
Update: Nope, that wasn't it.
Weird. I'm already running a CherryPy app in production on this server, so why won't this one start? I can't even get a trivial sample program to run. Same version of Python and CherryPy too. Hmm.Update: Foo. Networking wasn't running correctly on the newly-created virtual server. I restarted it, and everything is fine. It looked like a firewall issue - with the new system, not only is the server protected from the outside world, but all the virtual servers are protected from each other, and, in fact, from themselves. But no, it was just that the loopback interface wasn't running.
Update: Okay, looks like the application server and database are working fine now. Just need to set up the load-balancer. But it's nearly midnight, so I won't be doing the move today; it will be tomorrow instead.
Posted by: Pixy Misa at
11:42 PM
| Comments (5)
| Add Comment
| Trackbacks (Suck)
Post contains 423 words, total size 3 kb.
Saturday, May 23
Two Removes Are Equal To One Fire
Moving mu.nu into its new home right now.
I've reached the quiet spot between frantic preparation and even more frantic bug-fixing, where the files are doing their final update. Back to panic mode in 3... 2...
All finished! Not too much pain involved, either. Just waiting for the DNS changes to roll out now.
Moving mu.nu into its new home right now.
I've reached the quiet spot between frantic preparation and even more frantic bug-fixing, where the files are doing their final update. Back to panic mode in 3... 2...
All finished! Not too much pain involved, either. Just waiting for the DNS changes to roll out now.
Posted by: Pixy Misa at
04:36 PM
| Comments (15)
| Add Comment
| Trackbacks (Suck)
Post contains 61 words, total size 1 kb.
Saturday, February 07
1.2-alpha-c1
The new edition of Minx is working!
Okay, it's still in internal alpha testing, but after I ripped out its internals to provide for some of the major new features (scripting language, API support, safe third-party applets) it's been kind of broken for a while.
Now it's working fine, and seems to be at least close to the speed of 1.1. I can optimise the common cases of my new tag dictionary class, which is very heavily used; that should help things along.
Hmm.
Before:
After:
But I just noticed that recent comments aren't working on 1.2.*
Based on the work I've been doing at my day job over the past year, I do have some neat ways to potentially speed up the query side of things (and reduce the number of queries). But that's only one third the total time, so Amdahl's Law applies in spades.
* Doh! It's that blasted cross-referencing indexy thing I implemented. Efficient, yeah, but if the cross-reference table is empty it don't work. And if the cross-reference table isn't accurate, it don't work right. I'll have to find - or rewrite - the code to rebuild the index, but at least I know it's not a bug. As such.
The new edition of Minx is working!
Okay, it's still in internal alpha testing, but after I ripped out its internals to provide for some of the major new features (scripting language, API support, safe third-party applets) it's been kind of broken for a while.
Now it's working fine, and seems to be at least close to the speed of 1.1. I can optimise the common cases of my new tag dictionary class, which is very heavily used; that should help things along.
Hmm.
Before:
108kb generated in CPU 0.13, elapsed 0.1315 seconds.
65 queries taking 0.0324 seconds, 276 records returned.
Powered by Minx 1.1.2-beta.
After:
106kb generated in CPU 0.12, elapsed 0.1285 seconds.
70 queries taking 0.0426 seconds, 279 records returned.
Powered by Minx 1.2-alpha-c1.
But I just noticed that recent comments aren't working on 1.2.*
Based on the work I've been doing at my day job over the past year, I do have some neat ways to potentially speed up the query side of things (and reduce the number of queries). But that's only one third the total time, so Amdahl's Law applies in spades.
* Doh! It's that blasted cross-referencing indexy thing I implemented. Efficient, yeah, but if the cross-reference table is empty it don't work. And if the cross-reference table isn't accurate, it don't work right. I'll have to find - or rewrite - the code to rebuild the index, but at least I know it's not a bug. As such.
Posted by: Pixy Misa at
11:08 PM
| No Comments
| Add Comment
| Trackbacks (Suck)
Post contains 244 words, total size 2 kb.
<< Page 1 of 26 >>
85kb generated in 0.174 seconds; 60 queries returned 235 records.
Powered by Minx 1.1.4-pink.
Powered by Minx 1.1.4-pink.
