| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Comments:
<0> wait_timeout of 10
<1> % matches 0 or more characters
<2> redLED: How did you set the variable?
<0> i still have 1,300 active threads
<1> Use _ if you mean "1 character"
<0> GeneralFailure, set global wait_timeout = 10
<1> redLED: changing wait_timeout won't affect existing connections.
<0> restarted
<0> same problem
<2> It already has that many connections?
<0> nah
<0> around 100 so far
<0> but normal load for that machine
<0> is like less than 15
<0> it will get there shortly though
<0> allready at 200
<2> This may or may not help: Try setting the global var thread_cache_size to 200
<0> show status suggests there isnt an insane number of them being created
<0> plus, im kernel 2.6 with latest glib (centos 4), should have decent pthread etc...
<0> but let me try
<1> most likely you just have background operations, fragmentation or simply exceeded in-ram set bad enough to get disk bound
<1> but have fun
<0> this is the problem Therion
<0> my database
<0> is smaller than the ram available to me
<0> yet performance is truly horrible
<0> its a quad xeon 3ghz machine, running 64 bit centos 4
<0> 8gb ram
<0> mysql 5.0.4 atm
<2> Nice machine, I wish I had one of those
<1> It is idiotic to run 5.0.4
<0> and every time, about 10 mins after i restart the server and it gets load
<0> stuff breaks badly
<0> Therion, had exactly same problem with 4.1.x
<0> upgraded as act of despair
<1> So you upgraded to an alpha? :P
<1> Use 4.1.18 or 5.0.19
<1> Nothing else
<0> im in my 40th hour orso of debugging this :p
<0> i dont normally harras strangers on irc
<0> but this thing serves some 5.2 million users a day
<0> and is really pretty down :(
<2> redLED: When you restarted the server you lost your wait_timeout settings
<0> nah
<0> i stuck that in my.cnf
<0> and show innodb status, while might help me, in my case produces about 20 pages of the '-' sign, some data, followed by 20 pages more of the '-' sign
<2> Using the command line client?
<0> yes
<2> Use \G instead of ; at the end of the line
<2> It will be far more readable
<0> ah
<0> now there's a useful peice of advice i was unaware of :>
<3> hi
<3> has anyone ever used godaddy.com?
<3> the domain?
<2> bo0o: I have one, but only because it was free. And I still want my money back
<0> 8 queries inside InnoDB, 1318 queries in queue
<0> :(
<2> hmm. Sounds like a locking issue
<2> SELECT @@autocommit;
<3> Has anyone ever bought a domain from godaddy.com?
<0> GeneralFailure, it's 1
<2> redLED: Okay, that's good
<2> One sec, I think I may have an idea
<2> Let's try reducing InnoDB's lock wait timeout
<0> did that
<2> And no effect?
<0> innodb_lock_wait_timeout = 5
<0> none.
<2> innodb_table_locks?
<2> Just stumbling around in the dark here
<0> 1
<2> Try setting it to 0
<4> redLED you saying that paged produce '-' tells me that you have no idea where the probelm is
<4> bo0o no but 6 million other people have
<4> sorry 2.8 million
<2> bo0o: I already gave my opinion on GoDaddy above
<4> oh wait, 10 million
<0> firewire, if i *had* any idea where the problem is, i wouldnt be here in the first place.
<4> redLED let's try to break it down a bit
<2> The only reason GoDaddy is free is because they are cheap, but at the expense of ease of use
<4> first, take a deep breath (I'm serious)
<2> That didn't come out good. The only reason GoDaddy is popular is**
<4> GeneralFailure being hard to use doesn't make you money. godaddy is cheap because they make so much off of other services
<2> For domains, I prefer eNom -- you just have to find a cheap reseller
<5> Major Malfunction and General Failure
<0> im totally in for the deep breath
<4> redLED ok without making any ***umptions on where the problem is describe to me what you're seeing from a user standpoint
<2> Apachez: Who is General Failure and why is he reading my disk?
<0> ok
<0> i have a database, on a pretty big server, which i expect to handle a lot of web traffic (php stuff) sql queries
<4> ok
<0> that database is mysql innodb
<0> it did so fine, for a long time
<0> a few days ago, it stopped doing that, gradually at first
<4> what do you mean by stopped?
<0> until eventually it just got overloaded and system load crawled to max
<0> well
<0> first it would have high load
<0> get restarted
<0> and load would dissapear
<4> what kind of load
<4> cpu or disk
<0> system load
<0> cpu
<4> ok
<0> eventually
<0> no ammount of restarting helped
<4> can you log into mysql?
<0> yes i can
<6> is it possible to convert a mysql table column type from blob to text w/o losing the data?
<4> if you run show processlist are there users that are running queries for a long time?
<2> jmoschetti45: Theoretically, they are the same size
<4> jmoschetti45 yes
<6> so if it try it in phpmyadmin things should go ok?
<6> ok
<4> if you're worried take a backup first
<0> firewire, there are 2,500 processes total, ranging from 2500 to 1 for the time value
<0> firewire, it would appear quite a lot of them lean toward the 2500
<4> redLED ok do the top 10 look like they're running about the same query?
<0> the top 1 is composed out of 3 different queries on 3 different database
<0> top 10 even
<4> hmm
<0> but then again
<0> top 100 orso
<0> are all these 3 queries
<4> ok
<4> chances are they aren't very efficient
<4> are they select queries?
<2> MySQL, you've got some 'splaining to do
<0> while i am sure they arent (they are generated by 'pmachine' and 'drupal', 2 popular php cmses)
<0> i insist that since i can easily see the load on the webservers which generate these queries
<0> this same setup used to work just fine with no load, processing same number of queries till recently
<4> redLED first thing to do is kill them. Since they were kicked off by a web user that has long since moved on
<0> okay
<4> if they are select queries
<0> they are the top 100
<7> What kind of software should i use to "map" out a data base with table and such???
<4> ok
<0> are nearly all select queries
<4> go ahead and kill the highest running selects
<4> this will at least get your site back online
<0> just as a btw
<0> my entire top 100
<0> infact, all 2500 of them
<0> are a completely gradual ladder
<0> as in
<0> first one is 1500
<0> then 1499
<0> etc
Return to
#mysql
or
Go to some related
logs:
gentoo passwd: Critical error: immediate abort
#web
suse init_udevd_socket install
nozoki3
#debian
#ubuntu
billiards+ how to rack
debian x window troubleshooting no screen found
#web
#centos
|
|