Eustice: Routhier: emphazie “should” : I’ve never tested that, because all mysql servers I ever used went 64bit before we had more than 4GB RAM in them :
Stanowski: Groskreutz: ok, so 4 instances, each with 768 MB mem ?
Killion: Haris: should work, yes. check if the buffer pool size has to set to the *total* memory or to the size per instance
Lumb: Groskreutz: no such var in 5.1.73
Sagayaga: Haris: ah, correct. only since 5.5.4 😀
Routhier: Haris: you need to decrease innodb_buffer_pool_size to 1.4G or upgrade to 5.5 or 5.6
Routhier: Haris: it might work higher than 1.4G but this number sticks in memory for some reason
Shvey: My master is 5.0.95. my other slave is 5.1.69
Routhier: Maybe buffer pool overhead
Routhier: Haris: both versions are EOL
Routhier: Haris: promote you 5.5 slave that is running on x86_64
Resenz: Http://pastebin.ca/3172417
Routhier: Haris: you should also set innodb=force in my.cnf and/or sql_mode=no_engine_substition, which are both effectively on by default now
Routhier: Tldr, upgrade to 5.5 or 5.6.
Routhier: Groskreutz: that is tomorrow
Helmsing: Get a cloud machine as a slave promote the old slave to a master, running 5.5, and upgrade your 32bit boxes
Routhier: When the 5.5+ slave on 64 is promoted
Luebbert: Helmsing: i have it i hope https://dev.mysql.com/doc/refman/5.0/en/example-maximum-column-group.html
Falvey: Going to upgrade this one, thereafter holidays till the weekend
Sober: Looking for ways to bring it up. so at least its alive till that time
Mederos: 150924 0:25:14 ERROR Wrong option to sql-mode. Options given: no_engine_substition
Rasnick: Haris: reduce innodb buffer pool size on the slave substantially. ie 1500M
Chien: Its not even accepting 2G as well
Arizzi: Haris: exactly. try 1500M
Naslund: You should try the things people tell you
Stupak: Its working with 2G. please bear with me. trying a few things
Routhier: Routhier Haris: it might work higher than 1.4G but this number sticks in memory for some reason
Routhier: Haris: k. hopefully it doesnt go down in a couple hours after it seems ok.
Routhier: I would stick to 1200M or 1400M or so in your scenario, as i have said repeatedly
Furtak: Replicated temp table is building on it. will know in 30 or so mins
Holle: Replication stopped on this one because of mem issue
Secker: I have a db with all tables using auto-inc primary keys – I want to migrate to uuids instead, any thoughts on how I could do that while keeping relationships?
Warley: Kbond: why do you want to do this? it’s not a great idea . UUIDs use a lot more space than an auto-incrementing integer.
Maxin: Kolbe: I need to control the creation of IDs
Jindra: I need to know the id before inserting
Warley: Kbond: i know it’s annoying, but . why?
Moriera: Kolbe: i need to send the id back to the user – the insertion happens later
Warley: Kbond: you want to give some information back to the user even though nothing has been inserted into the database? that sounds very unwise
Warley: Or at least pretty weird
Gurganious: Kbond: do you need uuids in particular? Could you use a sequence table and a trigger?
Mccra: Kolbe: it is like a “promise”
Warley: Kbond: right, but you can’t promise anything to anyone unless it’s committed to the database
Hogrefe: Gurganious: interesting, i will look at that
Chodorov: Kolbe: if it fails, the user is notified later – i believe facebook does something similar – when you post a comment, it ***umes it worked and doesn’t wait for confirmation before showing it
Warley: Ok well then that’s different than a promise .