Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
MySQL upgrade

By hurstdog in Site News
Mon Feb 11, 2002 at 12:24:49 PM EST
Tags: Technology (all tags)
Technology

Well it seems we finally outgrew the stock mysql. Last week we started having a little problem with performance. After postulating and turning out to be wrong, we looked a little deeper. We found the problem.

Update [2002-2-11 13:12:20 by rusty]: Also, an apology about email...

Update [2002-2-12 14:53:43 by hurstdog]: We've disabled the whose_online box for the time being, its got a hell of a slow query in it, that we're seeing what we can do to fix. It was single handedly forcing the load of the db server up to about 20, from 1.5


It seems that mysql locks the whole table for an update. If you follow the last link above, it details a little problem that can occur under "Table locking is, however, not very good under the following senario." Kuro5hin got to that following scenario last week.

So we upgraded to MySQL Max 4.0, which allows for row level locking on updates and other queries. Also it gives us transactions, but we have no real need for that yet (scoop isn't set up to use them) so we have it set to autocommit after every insert/delete/update. So far it appears to be much faster. The load on the database server is significantly lower than it was last week around this time. We're still watching it and seeing if we need to tweak any last variables to get an extra ounce of speed out of it, but all seems to be well. Let us know here if you find any problems, or funny stuff starts happening.

In other news, those of you who are subscribed to Digests probably got five of them last night. I apologize; we were having some trouble with crons. Please don't report us to the spam police, it was an accident!

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Poll
K5 is...
o Faster than last week 48%
o Slower than last week 2%
o About the same 12%
o I wasn't here last week 36%

Votes: 248
Results | Other Polls

Related Links
o Scoop
o a little problem
o postulating
o problem
o MySQL Max 4.0
o Also by hurstdog


Display: Sort:
MySQL upgrade | 68 comments (68 topical, editorial, 0 hidden)
Questions (3.20 / 5) (#1)
by Vladinator on Mon Feb 11, 2002 at 12:27:27 PM EST

MySQL 4.0, is free or not free? PostGRE, will it work for you to solve this problem? So, DB was cause of problems, ja?
--
LRSE Hosting
Heh (5.00 / 2) (#2)
by spiralx on Mon Feb 11, 2002 at 12:30:26 PM EST

I was wondering who that other account was. Should've guessed it was you...

You're doomed, I'm doomed, we're all doomed for ice cream. - Bob Aboey
[ Parent ]

Hey yeah! (5.00 / 1) (#4)
by DesiredUsername on Mon Feb 11, 2002 at 12:33:42 PM EST

Scott William Lockwood.

Play 囲碁
[ Parent ]
No no (3.66 / 3) (#6)
by spiralx on Mon Feb 11, 2002 at 12:39:47 PM EST

This one, ne?.

You're doomed, I'm doomed, we're all doomed for ice cream. - Bob Aboey
[ Parent ]

Ja! (5.00 / 1) (#7)
by DesiredUsername on Mon Feb 11, 2002 at 12:41:18 PM EST

ViljamL == WilliamL(ockwood)

Play 囲碁
[ Parent ]
Doh, yes :) (5.00 / 1) (#10)
by spiralx on Mon Feb 11, 2002 at 12:46:22 PM EST

Goddamn Mondays ;)

You're doomed, I'm doomed, we're all doomed for ice cream. - Bob Aboey
[ Parent ]

Hahaha (5.00 / 1) (#11)
by Vladinator on Mon Feb 11, 2002 at 12:46:44 PM EST

I was amazed that no one figured it out before now. Even my wife was fooled, and she'd been reading it all day going "this sounds like someone I know, who is this?"
--
LRSE Hosting
[ Parent ]
Backwards (5.00 / 1) (#9)
by Vladinator on Mon Feb 11, 2002 at 12:45:44 PM EST

Actually, it's William Scott Lockwood III, not the other way around. :-)
--
LRSE Hosting
[ Parent ]
Oops. (3.25 / 4) (#8)
by Vladinator on Mon Feb 11, 2002 at 12:44:56 PM EST

Hehehe. Oh well. It was fun while it lasted.
--
LRSE Hosting
[ Parent ]
That's why I don't have multiple accounts (5.00 / 3) (#16)
by wiredog on Mon Feb 11, 2002 at 01:35:51 PM EST

Like Rusty with the Abooey account, and you with the Lockwood. It's too hard to keep them straight, especially on a Monday.

Peoples Front To Reunite Gondwanaland: "Stop the Laurasian Separatist Movement!"
[ Parent ]
Abooey (5.00 / 2) (#32)
by rusty on Mon Feb 11, 2002 at 05:39:12 PM EST

I am not Bob Abooey!
  • I own the site, so why would I need another account?
  • That diary pretending to be by me had several glaring inconsistencies.
  • You're just being led on by a clever prankster.


____
Not the real rusty
[ Parent ]
Clarification (3.66 / 3) (#39)
by Emperor Rusty on Mon Feb 11, 2002 at 07:21:50 PM EST

Astute subjects may become confused by our post above. While it is true that we do not play foolish games with multiple accounts, we do often use our priviledge to post alternately using our casual "nick" or our official name, depending on our mood. You can be assured that these names are attached internally the same account, though they may appear outwardly not to be.

And we would also like to again assure everyone that we would never impersonate our beloved court jester, even should that merry little prankster royally deserve it.
--------------------------------
Rusty I, Emperor of Kuro5hin, Protector of Scoop
[ Parent ]

Suuuure yer not! (5.00 / 2) (#40)
by wiredog on Mon Feb 11, 2002 at 07:37:07 PM EST

We believe you. Really, we do!

Peoples Front To Reunite Gondwanaland: "Stop the Laurasian Separatist Movement!"
[ Parent ]
To answer the actual question... (5.00 / 2) (#17)
by rusty on Mon Feb 11, 2002 at 01:36:09 PM EST

MySQL 4 is Free (GPL). Postgres would also probably have solved this problem, but it brings a lot of it's own for an application that was designed with mysql in mind. The PG port is still underway AFAIK.

____
Not the real rusty
[ Parent ]
AWOL (3.66 / 3) (#19)
by J'raxis on Mon Feb 11, 2002 at 02:03:32 PM EST

Your sig’s gone AWOL. Gotta love that Slashdot archiving, mm?

— The Raxis

[ J’raxis·Com | Liberty in your lifetime ]
[ Parent ]

I know (5.00 / 2) (#25)
by rusty on Mon Feb 11, 2002 at 03:47:01 PM EST

I'm leaving it, because it's embarrassing. Purposeful link rot is bad. If you're not gonna keep a URL working, don't bother making it.

____
Not the real rusty
[ Parent ]
uh (4.33 / 3) (#29)
by AnomymousCoward on Mon Feb 11, 2002 at 05:16:20 PM EST

"purposeful link rot is bad", yet your site does it all the time ....

try clicking on my user comment history, and viewing the very first (lowest rated, at the bottom) comment shown.

At the top of that page, there is a link that says: Car Lighting & Trade in Value, obviously to an old story. Click it.


QED.

Vobbo.com: video blogs made easy: point click smile
[ Parent ]
Um? (5.00 / 2) (#30)
by rusty on Mon Feb 11, 2002 at 05:33:35 PM EST

try clicking on my user comment history, and viewing the very first (lowest rated, at the bottom) comment shown.

Yep. Works fine. Glad we don't break links.

It's an editorial comment. If you can't see it, change your viewing prefs to "All Comments".

The second URL is a link to a story that was voted down. I see this as being a different thing than putting something up on a public area of the site and then expiring it. While the set of people who have access to it in the queue is large, it is still limited. It's not an open part of the site, and it should be clear that linking there is not wise. But I may just be justifying, feel free to disagree. Regardless of whether a story is posted or not, your comments are always available. And we've also gone to some trouble to make sure that URLs remain backward compatible, even after we changed the general URL scheme for the whole site, unlike some other organizations.

____
Not the real rusty
[ Parent ]

the second comment... (4.00 / 3) (#41)
by AnomymousCoward on Mon Feb 11, 2002 at 08:08:58 PM EST

The second URL is a link to a story that was voted down. I see this as being a different thing than putting something up on a public area of the site and then expiring it. While the set of people who have access to it in the queue is large, it is still limited.

If the story is voted down, why not remove the link entirely? Your comment that creating broken URLS is wrong is a valid point: go the extra mile and remove it. Your point that it is the victim of limited exposure is not: the discussion to which you link is a user created area, and as such, is not public either.

Vobbo.com: video blogs made easy: point click smile
[ Parent ]
I see (5.00 / 2) (#44)
by rusty on Mon Feb 11, 2002 at 09:40:20 PM EST

I see what you mean. Yeah, you're right. That link should only be there for people who can actually use it. Will add to bug list.

____
Not the real rusty
[ Parent ]
Been like that for a while, hasn't it? (3.25 / 4) (#3)
by DesiredUsername on Mon Feb 11, 2002 at 12:32:29 PM EST

For a DB that calls itself built for speed, I'm surprised that such a basic thing as row-level locking is missing.

OTOH, if you are measuring speed in ounces who knows what crazy ideas you'll think up next....8^)

Play 囲碁

What kernel you running? (none / 0) (#5)
by wiredog on Mon Feb 11, 2002 at 12:35:35 PM EST

From the mysql site:

Linux Advisory: Several of our users have reported random table corruption with Linux kernel 2.2.14 under heavy load. The table corruptions went away after an upgrade to a newer kernel. It appears that version 2.2.14 of the Linux kernel has a bug in the I/O implementation. Under some rather rare circumstances, the kernel will write incorrect data to the disk


Peoples Front To Reunite Gondwanaland: "Stop the Laurasian Separatist Movement!"
k5 runs 2.4.10-ac12 (none / 0) (#12)
by Captain_Tenille on Mon Feb 11, 2002 at 01:02:46 PM EST

So if there are problems with the kernel, that's not it.
----
/* You are not expected to understand this. */

Man Vs. Nature: The Road to Victory!
[ Parent ]

2.2.14 (none / 0) (#50)
by juahonen on Tue Feb 12, 2002 at 03:12:50 AM EST

Is this a MySQL problem or does anyone know of other applications for which 2.2.14 kernel causes data corruption. I have not heard of any other application with similar problems when running with 2.2.14 kernel.



[ Parent ]
maybe a little faster (none / 0) (#13)
by Delirium on Mon Feb 11, 2002 at 01:09:35 PM EST

But still kind of slow. According to Mozilla's timer it takes approximately 3 seconds to load any arbitrary page (this "post comment" page, any random story page, the diaries page, the moderate stories page, etc.). This is a little better than ~3.5-4 seconds last week, but not significantly. Significantly better than ~10-15s plus timeouts that it was a few weeks before that, but still on the whole slower than the ~1.5-2s load times back before the 4-week k5 downtime. So in my experience at least it seems that the new fancy multi-server redundant setup is still slower than Bubba was...

Faster than that here (none / 0) (#15)
by rusty on Mon Feb 11, 2002 at 01:25:45 PM EST

Using my always-reliable measure of "how many times does Galeon's foot pulse before the page starts loading" we're looking at about 3/4 of a pulse on average. Which is probably 1 to 1.5 seconds.

Are you counting from request to full render? Or just till you start getting data?

It's faster now than it ever was on bubba for me. And it's handling a much higher load now. What kind of connection are you on?

____
Not the real rusty
[ Parent ]

On my DS3 (none / 0) (#27)
by Miniluv on Mon Feb 11, 2002 at 04:20:01 PM EST

I just loaded the last three pages in less than .9 seconds (Moz0.9.8).

Requests though have been wildly variable in response time, I think K5 finally has a userbase where peak usage really makes a significant difference.

Some things are holy, and the sauna is one of them
[ Parent ]

Variability (none / 0) (#31)
by rusty on Mon Feb 11, 2002 at 05:37:10 PM EST

We've noticed the variable response time thing too. The currenmt theory is that it's basically caching-related. MySQL/InnoDB does a lot of caching, and so does Scoop, so early requests to a httpd/mysqld combination are slow. They quickly speed up though, as the various caches kick in. It's likely that the really slow requests are newly spawned pairs. There's not much we can do about it, unfortunately.

By the way, for faster indexes, turn off the "Who's Online" box in your display prefs. It only refreshes the list every two minutes, but when it does, that qury is dog-slow. Like 15 seconds slow. We're working on making it work better, but it's a tricky problem for something so unimportant. :-/

____
Not the real rusty
[ Parent ]

Don't forget (none / 0) (#35)
by Miniluv on Mon Feb 11, 2002 at 06:01:17 PM EST

If I'm the lucky bastard spawning a new httpd then I get all the overhead of perl startup, initial population of variables, etc. Hopefully you guys DO have maxrequests set on the mod_perl servers though, because I have a disgusting amount of firsthand experience hunting for mod_perl memory leaks, and it ain't fun.

Some things are holy, and the sauna is one of them
[ Parent ]
Memory leaks (none / 0) (#45)
by rusty on Mon Feb 11, 2002 at 09:45:31 PM EST

Actually, we have Apache::SizeLimit set up, which kills httpds when they get over (for us) 34Mb. The caching Scoop does makes it inevitably grow in size, since perl won't release memory once it has it allocated. So over time, any apache process will tend to grow. Usually they grow quickly up to a normal maximum, and then stay there for a long time. It's not a leak per se -- I also have long experience tracking those down, and to the best of my knowlege, Scoop is leak-free. But it's a similar sort of thing.

Under normal conditions, we shouldn't be spawning many new httpds.

____
Not the real rusty
[ Parent ]

wow (none / 0) (#49)
by boxed on Tue Feb 12, 2002 at 03:12:37 AM EST

If it were at all possible, a shift over to a java application server and a java codebase would totally remove the overhead of spawning new httpds and memory would be managed correctly. But of course, it's a way too big codebase to ever port over :/

[ Parent ]
JSP weblog (5.00 / 1) (#57)
by ennui on Tue Feb 12, 2002 at 12:08:40 PM EST

A team of Java developers and others interested in "weblog" software are currently working on a Scoop-like Collaborative Internet Media framework written to the JSP API. The core project is called Jism, an acronym for Java Interactive System for Media. The current model is about 70% compatible with the Scoop schemas to facilitate migrating Scoop sites to the Jism engine. It's uses JDBC so the backend database is basically transparent, and deploys well on Tomcat and Resin and should work on most application servers that support the JSP API.

A number of interesting features are being added to extend the basic functionality of weblogs like Scoop. In particular, a Wiki-style hypertext reference is tightly integrated so that a user-contributed knowledgebase can be built to complement the usual "storytelling" view a weblog generally presents, and has the pleasant side effect of allowing the simpler Wiki markup as an alternative to plain text or an allowed subset of HTML. Rich media is also supported, including a high-performance MPEG streaming servlet.

As the project progresses, the URIs of the alpha site will be announced (currently an NNTP interface has been prototyped in addition to the web interface) and make the "open source" portions of the code available. The Jism engine will most likely be released under the GPL or something similar.

"You can get a lot more done with a kind word and a gun, than with a kind word alone." -- Al Capone
[ Parent ]

got a url? (none / 0) (#64)
by boxed on Wed Feb 13, 2002 at 02:07:15 PM EST

and what framework does it use? or does it use it's own?

[ Parent ]
No public URL at the moment (none / 0) (#65)
by ennui on Thu Feb 14, 2002 at 09:52:47 AM EST

Portions of the project are hosted at sourceforge, but in light of recent devlopments with sf a new home is being sought. An announcement will be made when the alpha site is available.

"You can get a lot more done with a kind word and a gun, than with a kind word alone." -- Al Capone
[ Parent ]
Well (none / 0) (#58)
by rusty on Tue Feb 12, 2002 at 01:25:52 PM EST

Even as we speak, the mod_perl guys are working on this. Basically, perl was designed originally to be a short-run scripting language. That is, memory, once allocated, can be cleared and re-used by perl, but is never released back to the OS, because the language was only used for scripts that would run through and then exit completely.

mod_perl introduces a new longevity to perl "scripts" that makes this strategy less-than-optimal. I believe there are plans afoot already to allow the use of different garbage collectors in perl, depending on what kind of application you're writing. So a script might use the old style GC, while a mod_perl app could use a new GC that can release freed memory back to the OS.

____
Not the real rusty
[ Parent ]

Yep (none / 0) (#36)
by deefer on Mon Feb 11, 2002 at 06:05:07 PM EST

hink K5 finally has a userbase where peak usage really makes a significant difference.

K5 *burns* in the morning for us in the UK. Then about lunchtime (GMT), when the US wakes up, it slows drastically. But it *is* a lot quicker than it used to be - good work, fellas.


Kill the baddies.
Get the girl.
And save the entire planet.

[ Parent ]

faster now (none / 0) (#46)
by Delirium on Mon Feb 11, 2002 at 10:25:07 PM EST

For some reason it's in the 1.5-2s range now. I'm counting the time Mozilla displays in the taskbar ("Document: Done (1.932 secs)"), which I think is full render, but based on my highly accurate visual estimation at least 90% of that time is in waiting for a response (i.e. I click, wait 1-2 secs, then the page loads/renders almost instantly). I'm on a 10mbps switched university LAN connected through a Qwest ATM line. Pings report about 70ms consistently so I don't think it'd be network latency.

But anyway it's really fast right now. Usually it's about 30-40% slower though. It's not horrendously slow, but it's significantly slower than slashdot, for example (slashdot's speed is rather impressive though, especially considering what a mass of batshit Slash is). =P

[ Parent ]

Another Idea (5.00 / 2) (#14)
by LukeyBoy on Mon Feb 11, 2002 at 01:17:03 PM EST

Isn't MySQL Max an alpha branch of the code? That sounds like it has the features you want, but there might be an assload of problems in code that immature.

I know that the current vanilla branch of MySQL has support for the InnoDB type of table - which has single-row locking. Isn't this really the only feature from Max that you need? Migration from the MyISAM table to InnoDB might be a pain, but I think the standard version of MySQL is a lot more stable.

it is (5.00 / 1) (#22)
by hurstdog on Mon Feb 11, 2002 at 02:18:07 PM EST

I read all the docs I could find on it before I decided to use 4.0 instead of 3.0. The code isn't really that immature either, afaik, I think its more of a refactoring release, than a reimplementation. And we're making sure to keep backups, so if this starts acting all wonky, we should be able to revert back. *knock on wood*



[ Parent ]
Ah (none / 0) (#24)
by LukeyBoy on Mon Feb 11, 2002 at 03:35:02 PM EST

Thanks for the info. Sounds like I should check out Max for some of the sites I work with :-)

[ Parent ]
Alpha Code (none / 0) (#26)
by RadiantMatrix on Mon Feb 11, 2002 at 04:00:40 PM EST

Isn't MySQL Max an alpha branch of the code?
Perhaps, but so is Scoop.

--
No amount of genius can overcome a preoccupation with detail.

[ Parent ]
question (none / 0) (#18)
by streetlawyer on Mon Feb 11, 2002 at 01:43:28 PM EST

Was it this problem which led to my not getting an email notification for my "spindl3top" story which appears to have been hidden?

--
Just because things have been nonergodic so far, doesn't mean that they'll be nonergodic forever
related (none / 0) (#20)
by hurstdog on Mon Feb 11, 2002 at 02:13:16 PM EST

We thought that was the problem at first, so I threw up a patch that allowed us to disable that. So we disabled story emails, and that didn't fix it. So yesterday I re-enabled them, and as of about 11pm last night PST, it should be sending emails again. if it doesn't, let us know, cause then I put in a bug :-)



[ Parent ]
thanks (none / 0) (#21)
by streetlawyer on Mon Feb 11, 2002 at 02:15:19 PM EST

I thought that the conspiracy theory was the wrong one :)

--
Just because things have been nonergodic so far, doesn't mean that they'll be nonergodic forever
[ Parent ]
It is a conspiracy (5.00 / 2) (#23)
by wiredog on Mon Feb 11, 2002 at 02:50:19 PM EST

Just a different one than you thought it was.

Everything's a conspiracy.

Peoples Front To Reunite Gondwanaland: "Stop the Laurasian Separatist Movement!"
[ Parent ]

yeah (none / 0) (#47)
by tokage on Tue Feb 12, 2002 at 12:52:04 AM EST

And you're part of it. We're on to you, you goddamn stinking communist swine. You just better watch your back, that's all I can say.

meep

Pretending there is something left

is like pretending there was anything at all. - Angela Smith
[ Parent ]
You, yes YOU, tokage (none / 0) (#53)
by wiredog on Tue Feb 12, 2002 at 08:03:56 AM EST

Will be the one of first against the Wall when the Glorious Revolution™ comes!

Damn, we're gonna need a bigger wall...

Peoples Front To Reunite Gondwanaland: "Stop the Laurasian Separatist Movement!"
[ Parent ]

"It seems that mysql locks the whole table&qu (5.00 / 5) (#28)
by valency on Mon Feb 11, 2002 at 04:59:14 PM EST

Uh, isn't this why serious database applications use postgres or a commercial db?

I still can't figure out why people have such an obsession with mysql. What bothers me more is that so many people install mysql without knowing about this very serious limitation.



---
If you disagree, and somebody has already posted the exact rebuttal that you would use: moderate, don't post.
MySQL (3.33 / 3) (#33)
by rusty on Mon Feb 11, 2002 at 05:43:55 PM EST

K5 is not a serious database application. We have about two JOINs in the code, total. We have no transactions. We don't do any fancy database stuff. Mostly it's just supposed to be a fast data bucket.

MySQL has worked remarkably well until we started loading it with three or four fast Scoop servers all at once. Then the concurrent locking thing becomes a problem. Basically, we exceded the design specs of the database.

So far, InnoDB is working like a charm. It's much faster, and has handled the peak load today like a champ. I still can't figure out why people have such an obsession with Postgres, when mysql is easier to use and works just as well. ;-)

____
Not the real rusty
[ Parent ]

serious != query complexity (5.00 / 2) (#37)
by valency on Mon Feb 11, 2002 at 06:56:33 PM EST

K5 is not a serious database application. We have about two JOINs in the code, total. We have no transactions. We don't do any fancy database stuff. Mostly it's just supposed to be a fast data bucket.

IMHO, serious database application has far more to do with performance (specifically in highly-concurrent situations) than query syntax support.

Writing query parsers is easy, and the trivial implementation of the SQL command set is very simple to implement. Making it go fast (even on the simple queries) is the hard part ;)

I still can't figure out why people have such an obsession with Postgres, when mysql is easier to use and works just as well. ;-)

Transactions, scalability, ACID.

Actually, ever since they added BerkeleyDB tables (less than a year ago, I think), I started considering the combination of the two a "real" database. BerkeleyDB, despite not having SQL support, was always far more of a database than MySQL ever was to begin with ;) MySQL isn't much more than a query parser for BerkeleyDB in that sort of configuration ;)



---
If you disagree, and somebody has already posted the exact rebuttal that you would use: moderate, don't post.
[ Parent ]
InnoDB (none / 0) (#43)
by rusty on Mon Feb 11, 2002 at 09:39:20 PM EST

Transactions, scalability, ACID.

Check out the InnoDB stuff. It's nearly there.

____
Not the real rusty
[ Parent ]

Hehe (none / 0) (#66)
by Spendocrat on Thu Feb 14, 2002 at 03:34:27 PM EST

Worse is better!

[ Parent ]
Re: MySQL (4.00 / 1) (#48)
by juahonen on Tue Feb 12, 2002 at 03:04:18 AM EST

I would not trust a database which needs another DB to manage its own transactions. A RDBMS should be able to do its own transaction management without the aid of a third party database.

Makeshift support for basic RDBMS functions have plaqued MySQL all it's life. Sure it's fast., but I'd put more emphasis on reliability, crash recovery and SQL 92 support. Features like subqueries, foreign keys and stored procedures are coming to the following development release (4.1). It will take a long time until they are stable.



[ Parent ]
Hum hum (none / 0) (#67)
by Betcour on Fri Mar 01, 2002 at 10:35:55 AM EST

Actually there's no "third party database" : MySQL use a plugin architecture where table handlers are separate from the database engines. The default handler is MyISAM, which is robust but a bit plain (table locking, no transactions, etc...), and InnoDB is another table hander which support row-level locking, transactions and foreign keys. InnoDB is NOT a third party database.

[ Parent ]
Why PostgreSQL (5.00 / 1) (#52)
by WWWWolf on Tue Feb 12, 2002 at 07:35:18 AM EST

I still can't figure out why people have such an obsession with Postgres, when mysql is easier to use and works just as well. ;-)
Better features. And it scales better. And back in the day when I made the decision to use PostgreSQL, MySQL wasn't free software.

Also, "easier to use" is such a relative term. I'm still under the impression that the mySQL command-line client sucks... and on the SQL side, subqueries and other cool stuff that modern standards have brought to us are extremely nice and help to make things the easy way.

And the keyword is reliability. MySQL marketing: "It's fast, and everyone uses it.™" PostgreSQL marketing: "Acid-proof! A little bit slower than mySQL but twice as reliable! Inquire within for a mathematical proof!"

And it has somewhat appropriate mascot too. Dolphins are smart and fast, but elephants never forget.

-- Weyfour WWWWolf, a lupine technomancer from the cold north...


[ Parent ]
Postgres has functions (none / 0) (#54)
by billt on Tue Feb 12, 2002 at 10:06:35 AM EST

I still can't figure out why people have such an obsession with Postgres, when mysql is easier to use and works just as well. ;-)

For me in addition to the ACID stuff, it is functions. It makes the perl application alot smaller when you can execute multiple, conditional queries within the backend in one shot.

"Imagine how boring the world would be if all comment systems were as flawless as K5s?"--Rob Malda
--That comment has mysteriously disappeared!!



[ Parent ]
MySQL (none / 0) (#34)
by freija crescent on Mon Feb 11, 2002 at 05:57:04 PM EST

I used mySQL for a while. It was fast enough for my needs. Then I caught flak for using it, as it's not a true rdbms by account of those above me. They stated that it's inability to perform rollbacks, transactions, and row-locking made it essentially a glorified filesystem with an sql interface. I seem to see this a lot, such as on Slashdot, and several other websites that are DB intensive.

I wouldn't know though, i've since moved to informix. Maybe Max4.0 for mySQL makes up for this, but I'm kind of interested in how the different dbm systems out there work internally, and which is the more scalable, secure, etc.

-fc

I wasn't here last week (none / 0) (#38)
by X-Nc on Mon Feb 11, 2002 at 07:20:22 PM EST

I happened to be in Thailand for the week. While I did check email every day I didn't actually do anything else. They have Internet Cafe's there that are very inexpensive and resionably fast (if you get on after 7pm local time).

There was a very good reason I didn't do anything more than email. Her name is Sri. Man, I think I'll have to get off my @$$ and add some diary entries.

--
Aaahhhh!!!! My K5 subscription expired. Now I can't spell anymore.

Coming from Slashdot (none / 0) (#68)
by Silent Chris on Mon Mar 04, 2002 at 04:28:31 PM EST

I wasn't here last week because I came from Slashdot. Once I read they were going to bigger and more obtrusive ads, I decided to finally make the switch. Hey guys. :)

[ Parent ]
New Scoop bug (none / 0) (#42)
by wiredog on Mon Feb 11, 2002 at 08:14:50 PM EST

When I make a comment, and then hit the back button to go to the main page, I get the comment I just entered instead of the main page. Happens in Konq and IE.

Oh, and word wrap doesn't seem to work in Konq...

Peoples Front To Reunite Gondwanaland: "Stop the Laurasian Separatist Movement!"

More about word wrap in Konq (none / 0) (#51)
by Ranieri on Tue Feb 12, 2002 at 03:19:51 AM EST

The edit box seems to wrap text at about two times its width, forcing me either to scroll or to include enters manually to keep it readable. Fortunately I use HTML mode, and the result isn't nearly as irritating as it would be if I were using ASCII mode.
Furthermore sometimes when I reload the the page (or after a preview) the text seems to be only ten or so characters wide. This promptly changes back to normal if I click on it.
Apparently the edit box has a profound identity crisis, especially regarding its width.

While this is not a major bug, it is quite irritating. I would appreciate it if someone could look at it when all the vital problems have been sorted out.
--
Taste cold steel, feeble cannon restraint rope!
[ Parent ]

Posting in Konqueror... (none / 0) (#59)
by rusty on Tue Feb 12, 2002 at 02:26:55 PM EST

Well, here I am posting in Konqueror to see what the problem might be. Hmm, the line just wrapped, and it did it properly. This is version 2.2.1 by the way. Yup, it definitely wraps correctly here. Does it make any difference if I use plain text mode? Let's preview and see what happens. No, no difference.

What version of Konqueror is this? I'm temped to call it a browser bug, since I can't seem to replicate it.

____
Not the real rusty
[ Parent ]
Possibly due to variable vs. width fonts (none / 0) (#60)
by hillct on Tue Feb 12, 2002 at 05:14:33 PM EST

I'm guessing here but it's possible conquerer counds the chars perline to determine where to wrap, rather than the pixels so if you use a variable width font in the browser display, you would get odd results as to where wrapping occurs. I don't use or have Konqueror installed so I havn't tested this theory.

--CTH


--Got Lists? | Top 31 Signs Your Spouse Is A Spy
[ Parent ]
I have the same problem. (none / 0) (#61)
by special ed on Tue Feb 12, 2002 at 06:20:24 PM EST

I am also running 2.2.1 on KDE 2.2.2. However, I think it might be a problem with Konqueror, since I have the same problem with a couple other sites.

Meanwhile, the world turns foolishly on and ants tickle his butt.
[ Parent ]
More info on Konqueror (none / 0) (#62)
by Ranieri on Wed Feb 13, 2002 at 04:27:24 AM EST

I'm using Konq 2.2.2 on KDE 2.2.2-2.
Changing font does not seem to alleviate this problem, i tried all the fixed size fonts i have (including adobe-courier and the ttf courier new). I remember it working correctly at first, then breaking. This could have happened around the time i upgraded KDE/Konq. It might therefore not be unreasonable to regard it as a Konq 2.2.2 bug until proven otherwise.

I'm sorry to have wasted your time on it, rusty. I'll have a look on the Konq mailing lists to see if it's a known problem, and then i'll go bug them about it :)
--
Taste cold steel, feeble cannon restraint rope!
[ Parent ]

Not a Scoop bug (none / 0) (#63)
by Ranieri on Wed Feb 13, 2002 at 04:37:53 AM EST

This problem also shows up on ``the other site'', and apparently it is known.
--
Taste cold steel, feeble cannon restraint rope!
[ Parent ]
Speaking of new bugs (none / 0) (#55)
by mauftarkie on Tue Feb 12, 2002 at 10:26:51 AM EST

Maybe I missed something, or maybe you already know about it -- in which case I apologize in advance.

I read this story yesterday. Today, my front page said there were 14 new comments. I click the story... and no "[new]" text marking the new comments at all! It's also very inconsistant -- most of the time I see the "[new]" text in the new comments, but every so often I don't...

This started maybe 2-3 weeks ago. I started noticing it on the Trouble over at [the other site] story.

Lastly, am I supposed to head over to scoop.k.o to file bugs, or ... ? I didn't see anything in the FAQ here. I want to do the right thing, of course.

Thanks!


--
Without you I'm one step closer to happiness without violence.
Without you I'm one step closer to innocence without consequence.


[ Parent ]
we know about it (none / 0) (#56)
by hurstdog on Tue Feb 12, 2002 at 10:57:23 AM EST

Generally posting bugs to your diary or to site news stories ensures that we will see them. I do searches for 'scoop' every once in a while to see if I missed anything in my reading of the site. The bug you described is due to some caching that we do, and we're working on it. Thanks for the notification though.



[ Parent ]
MySQL upgrade | 68 comments (68 topical, 0 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!