#mysql
18 October 2007
Total 25 pages. You are browsing page 3/25.
First :: Prev :: [1] [2] [3] [4] [5] [...] :: Next :: Last
02:45 <****> I'm suspecting a bug
02:45 <****> The state is "Sending data"
02:45 <****> But it's clearly not :/
02:46 <****> I had queries stuck in "Sending data" once.
02:46 <****> but I don't know what to ask, its more the concept that I don't get basicly i have tables like recipe, ingredientlist and ingredients and I need to use the recipeid in the recipe table to compare to an ingredientid in the ingredentlist to get the ingredients from the ingredient table but I dont understand how to even do that
02:46 <****> I had assumed it was just really slow, and had meant to terminate it.
02:46 <****> But Ctrl-C exited the client and didn't terminate it.
02:47 <****> (The query then crashed my entire site when it prevented my backup script from taking out a lock on an obscure and fairly worthless table, for twelve hours.)
02:47 <****> Whether this is of any help to you, I have no idea. :D
02:48 <****> Simetrical: What version was that ?
02:48 <****> And was it repeatedly doing it ?
02:48 <****> This is more or less a nightly occurance for us
02:49 <****> clergyman, it's repeatable. Version is 5.0.22-log. It was a one-time query, so it didn't actually repeat, but it seems to exhibit the same behavior if I repeat the same query.
02:49 <****> clergyman, this was a somewhat convoluted SELECT with subqueries. I didn't see any reason why it would take so long to execute, since each subquery was pretty fast.
02:50 <****> Simetrical: Run EXPLAIN on it
02:50 <****> seekwill, bah, I have to look it up.
02:50 <****> Simetrical: Joining derived tables can behave rather badly.
02:50 <****> No, no joining, but it was unions.
02:51 <****> Simetrical: union of two simple selects?
02:52 <****> It was a horrible query, I was trying to get some statistics for a one-off and just threw together whatever looked like it might work even if it would take a couple minutes.
02:52 <****> http://pastebin.ca/740572
02:53 <****> Better: http://pastebin.ca/740574
02:54 <****> That illustrates the "hanging on 'Sending data'" behavior.
02:54 <****> That was just eleven seconds, but it went for nearly twelve hours last time I (accidentally) let it.
02:54 <****> Have you raised a bug report ?
02:55 <****> No, I just thought it was slower than I expected.
02:55 <****> Since we can't reproduce it and there's absolutely fsck all useful things in processlist and innodb status etc. there's not a whole lot to report on our behalf :
02:55 <****> What happened when you killed it? Did it die ?
02:55 <****> And actually got out of the queue
02:56 <****> clergyman, yes.
02:57 <****> If I go ALTER TABLE lineItems ADD UNIQUE KEY lineId (transactionId, lineNumber); ....
02:57 <****> Also, if I run it with only one of the subqueries (no UNION), it takes 0.88s. If I run it with two, it takes at least five or ten seconds.
02:57 <****> Before I kill it.
02:57 <****> Will it make sure that only the combination of the two keys is unique?
02:57 <****> Simetrical: Ah, ours doesn't
02:57 <****> Or that both are individually unique
02:57 <****> Arcturius, that the combination is unique.
02:57 <****> Arcturius, if you want individually unique, put one unique key on each.
02:58 <****> Simetrical: Yeah, I was trying for a combination. I just wanted to make sure I was correct.
02:58 <****> clergyman, makes our problem pretty different, then.
02:58 <****> (Not that mine is really a problem.)
02:59 <****> Yeah :/
02:59 <****> I have yet to bump into someone that's seen this issue or know what to do about it
03:00 <****> Core dump!!!
03:00 <****> It's not crashing
03:00 <****> It's happily sitting there with this query in the queue, doing aaabsolutely nothing
03:00 <****> You can still get a core dump.
03:01 <****> Really? How?
03:01 <****> kill -9 will normally do it on Unix, IIRC (never needed one). There are utilities to get one without killing it, too, I think.
03:02 <****> In my experience, kill -9 creates a core dump file in the cwd that I have to delete so it doesn't clutter stuff.
03:02 <****> I think you will need to specify that you want dumps at build time, or am I mistaken ?
03:03 <****> I kill -9 our mysqld and mysqld_safe almost nightly, and have never seen a dump
03:03 <****> Build time of what? Core dumps are a feature of the OS, not the app . . .
03:03 <****> clergyman: doing that can corrupt data
03:03 <****> kimseong: I know, it's extremely uncomfortable, but I just don't know how to fix it
03:04 <****> and no one else knows or have seen this issue
03:04 <****> It's not like I haven't spent hours or even days on this
03:04 <****> clergyman, ah, kill -3 apparently is what gives core dumps.
03:04 <****> Not kill -9. I think.
03:05 <****> Simetrical: Ok, I'll give that a go next time (tomorrow night :/)
03:05 <****> gcore apparently gets core dumps of processes without killing them,.
03:05 <****> .
03:05 <****> If available.
03:05 <****> Simetrical: I think if you compile some things with debug support, the crash handler rais()es 3 to give a core dump
03:06 <****> clergyman: 'I kill -9 our mysqld and mysqld_safe almost nightly' wow!
03:06 <****> Yeah :/
03:06 <****> Oh, ulimit -c apparently says whether core dumps are enabled.
03:06 <****> And what the max size is.
03:06 <****> They're disabled on my home PC and my server, evidently.
03:07 <****> Hm, on this server as well
03:08 <****> http://aplawrence.com/Linux/limit_core_files.html
03:11 <****> Thanks
03:11 <****> That was just a Google hit.
03:11 <****> Next step is making any sense out of the core :)
03:11 <****> That's the tricky bit.
03:11 <****> I've wrestled gdb before
03:12 <****> hey guys, i have a table with columns that might have nothing in it, but when i do select where column like '%' it still shows up, how can i make sure it doesn't?
03:12 <****> And my limited programming skills doesn't help either I guess
03:12 <****> Smidget, WHERE column != '';?
03:13 <****> a-ha, i'm dumb, thanks
04:28 <****> darn it. is 5.1.22 RC stable ?
04:29 <****> RC not GA yet
04:29 <****> yeah i just said RC
04:30 <****> oh i have 5.1.11
04:34 <****> anyone have a good link to a solid mysql setup tutorial? everything i see online is rather bad...
04:35 <****> ruz322: what kind of setup do you have in mind?
04:36 <****> !man install
04:36 <****> (Choosing an Install Type) : http://dev.mysql.com/doc/mysql/en/mysql-install-wizard-install-type.html
04:36 <****> !man installation
04:36 <****> (Installation Layouts) : http://dev.mysql.com/doc/mysql/en/installation-layouts.html
04:36 <****> yeah, better not wait more than one second for an answer.
04:36 <****> what's up with archivist?
04:36 <****> holiday?
04:37 <****> probly just out setting up a new social order... OH ! you said archivist [sorry, i had to say ONE bad joke today....]
04:39 <****> any one have an example my.conf file online that I could use?
04:41 <****> TheCougar: mysql installation provides you sample my.cnf
04:41 <****> hum where does it put it? i'm using the fbsd ports
04:41 <****> didn't see one yet
04:42 <****> support-files? or basedir
05:14 <****> hi
Total 25 pages. You are browsing page 3/25.
First :: Prev :: [1] [2] [3] [4] [5] [...] :: Next :: Last