| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Comments:
<0> HAH ****A
<1> LETS SEE HOW YOU HELP HIM IF YOU CANT SPEAK
<0> Abobo: there are many different ways to distribute 'load' depending on the workload of your application
<1> hahah
<0> ****a
<0> find
<0> fine
<2> mostly the load I refer to is mysql's calculation load
<0> I'm going to bed
<3> will: would this work, DELETE FROM t2 USING t1 WHERE t2.id <> t1.id
<2> large calculations and such
<0> Abobo: write (insert, update, delete) or read (select)
<1> Colloid, No.
<2> mostly read
<0> Abobo: replication is well suited to distribute read load
<2> problem is, with replication both servers can't write, right?
<3> Will: what's wrong with it?
<4> firewire: very much appreciated thanks ... http://pastebin.com/645646
<0> Abobo: yes but you'ret trying to distribute read load
<2> the issue is, a server that calculates is also a server used by clients, which means it has to write
<1> Colloid, Everything?
<2> so I guess what I need is a way to distribute read and write load to many servers
<0> babo: where are the teachers?
<1> Colloid, Look in the wiki in topic. Go to regulars -> will. There is a sample SELECT query you'll need. You can use that as a template to create the DELETE
<4> firewire: oops ... sorry just one sec
<0> Abobo: so you want to setup your application to always write to one server and read from another server
<1> firewire, The teachers are in the cl***room. DUH
<0> or read from a load balancer infront of several servers
<0> will: the teachers are in the teacher lounge smoking, we all know that
<3> thanks
<2> yea, but that pretty much involves rewriting every update/delete/insert command in the entire application to route all updates to a different server, right?
<2> additionally, the user may not see an instant change, right?
<0> Abobo: no, it involves rewriting the mysql_query calls to either go to a write connection or a read connection
<4> firewire: I need the search to be able to correspond to school_jobs too .. http://pastebin.com/645647
<2> thats pretty much what I mean
<0> mysql replication in normal conditions is usually up to date sub second
<2> I see
<2> and what if the master goes down?
<0> that's a whole different story
<3> will: DELETE FROM a LEFT JOIN b ON a.id = b.aid WHERE b.aid IS NULL
<1> Colloid, WOW. You impress me.
<2> I was told clusters is a better way to handle mysql replication, but cant find any info on that
<3> :<
<0> for now you want a very fault tolerant master
<2> I see
<3> will, will that work?
<1> Colloid, I'm totally serious. That looks right.
<3> <3
<1> Let me check to make sure.
<0> Abobo: clusters are good for distributing write load, not read load
<0> don't discredit mysql replication, it makes some very impressive things possible
<2> I totally agree, just trying to think of any faults
<2> basically I need a master to never fai
<2> l
<0> there are faults :)
<2> and then I have to rewrite the application to send update/insert/delete to master
<0> There are methods to failover a master
<2> if master is a remote server, I'm ***uming there will be connection delays
<5> I like that 'I need high availability at low costs' attitude
<0> you can also go the cl***ic dual master route
<2> hows that work?
<4> firewire: I think I could probably do it if I was able to use CREATE VIEW ... but I'm on mysql 4.x ... so it's not available
<3> ty will, u r a qt pi~
<1> Colloid, You need to add something after DELETE. Tell it what to delete.
<1> You scare me now...
<3> what do you mean?
<1> Re-read the DELETE syntax. Or just try it :)
<3> k
<0> babo: so you're looking for rows where teacher id is the same in teachers, and recuiter_jobs but not sent_matches?
<2> is it possible to write to a slave server? just not a good idea right?
<3> it should work, it'll just delete a row
<0> Abobo: I don't hink you need to
<2> what I mean is, I may want to replicate only select tables
<6> Abobo: possible, but you should not unless it is to a table or db which is nore replicated
<2> and stupid little things like cache tables, keep on slave
<0> Abobo: can you give us an overview of your situation?
<2> I do creditcard processing, we're getting big enough that a ton of transactions per minute is round the corner
<2> in addition, merchants and admins use the same servers. calculating invoices and payroll causes high load. I want to do that on a different server
<4> firewire: hmmm ... basically whenever I send a job to a teacher, I store the teacher_id and job_id in sent_match. I need a function that will pull out job matches that haven't already been sent and so don't exist in pairs in the sent_match table ...
<0> ok how many credit card transactions per second are you doing?
<2> so transaction server should probably be a slave, merchant/admin another slave, and then an independant master, right?
<0> babo: inner join teachers, school_jobs left join sent_match where job_id is null
<2> we're not there yet, average 10 thousand per day
<0> your payroll calculations can easily run off a slave
<0> same with invoices
<2> yep, so basically it should be as I just mentioned right?
<2> one for transactions, one for admin/merchant, and then an independant master?
<0> no
<2> by independant, I mean not public
<0> you want a single master for active transactions then a slave for reporting
<4> firewire: but the problem with that is if it only returns where job id is null, then if I send one job to a teacher, that teachers_id won't be null ... and so the teacher will be excluded from all other matches as well ...
<2> mmm, ok
<4> firewire: if you get what i mean ...
<2> is there any limit to number of slaves that can be used?
<0> Abobo: not anything you're going to hit any time soon
<0> 10,000 transactions per day isn't much :)
<2> gotcha, that should work
<2> yea, its not
<2> i'm planning for the future
<0> that's good
<0> you should go to the user conference this summer
<2> I dont want to be there and not be ready
<2> yea? where at
<4> firewire: how do I inner join teachers and school jobs ... inner join them on what ?
<0> santa clara, ca
<7> Hi All :-)
<2> i just might
<0> babo: do you know the difference between an inner and outer join?
<0> Abobo: yahoo is going to do a presentation on mysql fault tolerance
<2> so the rest should be technical right? if the slave fails, I could fall over to the master and its really just code that ensures its seemless
<2> nice
<0> it's good stuff, trust me
<4> firewire: yes, an inner join needs to join ON something though right ?
<0> babo: no
<4> babo -> w3schools ...
<2> for a quick fix, would you recommend writing a custom php function for mysql_query that checks the query syntax and determines whether to read or write? or is that a potential problem?
<0> if you inner join table a and table b mysql will only return rows where the condition matches a and b
<0> with a left join mysql will return rows from a and null for b where the condition didn't mastch
<2> unless you put a condition for b in the WHERE
<0> Abobo: I would recommend explicitly specifying the mysql connection for read or write for each mysql_query call and setting your slave to read-only so it catches any attempted writes
<2> you make a good case
<2> I appreciate the help. thanks
<0> no problem
<7> anybody help me with php conection to mysql and sets variables for this connection... my example for this:
<7> when i conect to mysql variables sets from section [client] in my.cnf? that's right?
<0> no
<0> php sets its variables from php.ini
<7> ?
<1> Well, usually you would specify connection settings in the PHP script itself...
<1> would/should
<0> php won't read the my.cnf files
<7> but php use mysql libraries for connect to mysql and its lib use my.cnf ;-)
<4> firewire: select * from inner join teachers, school_jobs left join sent_match where sent_match.job_id is null ... gives me a syntax error
<4> oops ...
<1> muhlik, No, the client reads my.cnf, not the libraries.
<0> babo: free hand queries will do that
<1> babo, No commas in that FROM!
<4> firewire: sorry, I meant SELECT *
<4> FROM teachers
<4> INNER JOIN school_jobs
<4> LEFT JOIN sent_match
<4> WHERE sent_match.job_id IS NULL
<6> missing join condition for the left join
<0> I'm sure you can resolve the syntax issues
<1> lokus, Also for the INNER :)
<0> I'm not here to be perfect
<4> firewire: ok cool, thanks ...
<4> lokus: the inner join doesn't need an ON apparently
<4> but the LEFT JOIN does ... ?
<6> then the inner becomes a cartesian product
Return to
#mysql
or
Go to some related
logs:
obontu
killingyourfather
firmware ubuntu f506
modprobing in ubuntu
a8e abbreviate
gaim undernet timeout
#linux
#math
gimp tutorial tree
#perl
|
|