| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10
<dammit> you'd just have to put all db independent caches into a separate cache domain
<dammit> eh
<will> dammit: For example, USE a; SELECT COUNT(*) FROM a.table; USE b; SELECT COUNT(*) FROM a.table;
<dammit> *query caches
<will> But it's not DB indepenednet. It's the SAME DB
<xlx> will, exactly.
<dammit> will: I fully understand
<will> dammit: Why should it matter which DB you're "in"?
<dammit> will: because you have to evaluate are there local references or not
<dammit> if there are local references
<will> hmm
<dammit> then you have to use db environment.
<Xgc> Because, (not in MySQL) it could be possible for X.tablename to refer to a link/alias that could point anywhere.
<dammit> will: query cache uses query string for cache...
<Xgc> It would be nice if MySQL supported something like that.
<dammit> then in cache it also has list of tables
<will> DB specific?
<Xgc> X means something different depending on which database you are currently connected to.
<will> So what if you did not specify which DB to default to?
<Xgc> s/connected to/using.
<will> Ommited using "USE" or mysql_select_db() ?
<will> Since those are optional.
<dammit> 1. Query_cache object consists of
<dammit> - query cache memory pool (cache)
<dammit> - queries hash (queries)
<dammit> - tables hash (tables)
<dammit> - list of blocks ordered as they allocated in memory
<dammit> (first_block)
<dammit> - list of queries block (queries_blocks)
<dammit> - list of used tables (tables_blocks)
<dammit> eh
<will> tables hash?
<dammit> for example, there're other issues
<will> xlx: TEST IT! :)
<dammit> regarding query cache
<dammit> like... character sets of connection
<dammit> damn
<dammit> at the start of that file
<dammit> there're 10 pages
<dammit> of documentation
<dammit> /* Key is query + database + flag */
<dammit> if you did not use a database
<dammit> else
<dammit> {
<dammit> DBUG_PRINT("qcache", ("No active database"));
<dammit> }
<dammit> sorry
<dammit> your query is not cached
<will> hmm
<will> Not sure about the logic behind that./
<dammit> oh wait
<dammit> checking more
<dammit> I'm after few beers
<will> :)
<dammit> so not that accurate
<dammit> ;-)
<dammit> anyway, if there is database, it is added to the key
<dammit> if there's no active database it is not.
<dammit> so if all table references are fully qualified, one could omit the database specification in key...
<dammit> This is what our Great Source tells us!
<will> But wasn't xlx's fully qualified?
<alnr> I am getting an error which says that more info is in "MySQL error log". where would that be on a debian sys? The file /var/log/mysql.err has 0 bytes.
<dammit> will: 'could' I mean 'one could patch it'
<dammit> right now there's no such functionality
<will> oh
<dammit> I've just reviewed the code
<dammit> if you want to fight me
<will> But we probably should find out WHY they require the DB specification in the key... I bet they have a good reason!
<dammit> DO IT NOW
<dammit> will: because query itself is a key
<will> No, I'm not arguing what MySQL does... I'm saying what SHOULD.
<dammit> not parsed one
<will> ah
<dammit> well, I did tell you like 10 minutes ago
<dammit> that MySQL could do this.
<will> heh
<dammit> if anyone wrote the code for it
<will> So MySQL checks qcache before really ever looking into the query.
<dammit> yes
<will> So it's a speed issue.
<dammit> this is why query cache is fast.
<dammit> it is not parsed
<will> xlx: That's why they do it!
<dammit> they would have to parse whole query in order to decide, are there local references or not.
<dammit> see, now they have hashes of updated tables
<dammit> and they have used tables attached to each query cache entry
<dammit> so that when you issue a query, it's located immediately in hash, then table modifications are checked
<dammit> (or it even invalidates queries on table modifications..)
<will> ok
<dammit> AT LAST
<will> :)
<dammit> ok, got to sleep a bit
<will> No no
<will> Stay up
<alnr> I am getting an error, "1201 (HY000): Could not initialize master info structure;" on start slave, which says that more info is in "MySQL error log". where would that be on a debian sys? The file /var/log/mysql.err has 0 bytes.
<will> Or else you're going to get a hangover.
<will> alnr: /var/lib/mysql
<dammit> alnr: RESET SLAVE;
<dammit> !tell us about disclaimer
<SQL> dammit asked me to tell you this: We are in no way responsible for the data you will delete, proceed at your own risk and know that backups save jobs.
<dammit> ;-)
<will> ;0
<will> :)
<alnr> hmm other than being unable to connect to master (those the master host 3306 telnets ok), that seemed to clear the problem
<alnr> thanks dammit
<alnr> thanks will but there doesnt seem to be anything error logish in /var/lib/mysql
<bigfoot-> Maybe try /var/log/mysql.
<will> Check your my.cnf file to see if it's placed somewhere else.
<alnr> bigfoot-: that seems only to have bin logs and there is a /var/log/mysql.err but it is empty. will check my.cnf
<Zalamander> how can I filter output in a "describe table" query?
<alnr> i mean, *I'll* check it
<Zalamander> For example, I'd like to do something like this syntactically invalid statement suggests: SELECT Field FROM (DESCRIBE some_table) AS something WHERE Key = 'PRI'
<will> Zalamander: Use MySQL 5.0
<dammit> Zalamander: use 5.0
<will> INFORMATION SCHEMA
<dammit> (I still stay up :( ..
<alnr> i guess reset slave goes along with recopying the db frm master to slave? or is there a way to know the last pos at which the slave which was reset left off?
<Zalamander> heh thanks for that tip. Any way to do that inside a query in MySQL 4.0 ?
<will> If there was, why would MySQL implement that in 5.0?
<will> If there was, why would we have suggested it?
<Zalamander> I ***ure you that your motivations are utterly unfathomable to me :-)
<dammit> alnr: well, if you already did reset slave, there's no way finding that out
<dammit> ;-D
<alnr> dammit: a warning bell went off before i did th reset slave, but I didnt listen :(
<Rub3X> can someone help me with an install =|
<dammit> alnr: usually you can simply take a look at the file
<dammit> it has all positions
<dammit> then you reset slave; change master with proper positions
<dammit> and here you go
<lawnchair> sql-bench is bull****
<lawnchair> and thats all i have ot say about that
<Rub3X> Force admin # mysqld_safe --user=mysql Starting mysqld daemon with databases from /var/lib/mysql
<Rub3X> STOPPING server from pid file /var/run/mysqld/mysqld.pid
<Rub3X> 051023 03:49:43 mysqld ended
<Rub3X> =\
<alnr> im trying 'load data from master' (instead of mysqlhotcopy for a change) but it ends too fast and seems to do nothing
<will> Rub3X: Use the Force.
<Rub3X> uh huh
<blazebud> what are the limitations to using a NDB cluster? i hear that GROUP BY doesn't work, what else?
Return to mysql
or
Go to some related
logs:
politics
JavaScript prefetch webcam
|
|