|
Share Bear posted:when would you use postgres instead of mysql besides hating on the man, the myth, the legend when you need to do more than just shove data in tables and then retrieve the exact same thing later. or you need to use an index more complex than a b-tree or hash. or you actually care about transaction support for more than crud operations. or you sort of care about preserving the integrity of your database and not being able to do poo poo like deleting the tables behind your views without your db having a problem with that i can't say that the mysql c client barfing out gigabytes of warnings at me the last time i had to compile it really helped my perception either i'm also gonna say db2 is the best even though i've never used it or read anything about it since i really like mainframes and that kind of fancy hardware crap
|
# ¿ May 12, 2016 23:04 |
|
|
# ¿ Apr 28, 2024 21:13 |
|
mysql has a query optimizer?
|
# ¿ May 12, 2016 23:24 |
|
i mean, other than the dude you pay to rewrite queries all day
|
# ¿ May 12, 2016 23:25 |
|
qhat posted:the best thing about mysql is the stored procedures don't have their query plans cached server side. they are generated every single time a connection is opened. conversely, i've seen a mysql server holding onto huge quantities of memory and just generally making GBS threads itself after an application generated a whole lot of prepared statements by mistake. even after the connection had been closed and associated resources should have been cleaned up. never figured what that was all about and the server had to be rebooted in the end
|
# ¿ May 12, 2016 23:37 |
|
MALE SHOEGAZE posted:why is it even called mysql named after one of the creator's daughter i think
|
# ¿ May 13, 2016 20:27 |
|
have they fixed the thing where doing a write acquires an exclusive lock on the entire loving database, making write performance utter trash and generating contention with the rebalancing process thus ensuring that adding a new machine to the cluster will take on the order of a week if the db is under load?
|
# ¿ May 14, 2016 16:45 |
|
huh. that's kinda unexpected, i assumed mongo would just keep using whatever they had before while proudly advertising "web-scale read/write scalability!" on their front page
|
# ¿ May 16, 2016 15:47 |
|
didn't you mention shitcanning a 45k "dba" not too long ago? because i think the rest of them could use the same treatment. being able to do backups of the db you maintain is kind of a pretty basic+important skill imo
|
# ¿ May 17, 2016 18:58 |
|
zen death robot posted:I didn't post the Nonstop SQL thing as a joke, we actually have some customers that demand that kinda poo poo is nonstop sql distributed across datacenters? because my favorite in that category is google's f1 which is a sql db with synchronous multi-datacenter replication they use for their ads backend "oh, the us west coast just fell into the sea? whatever, no biggie, we got 3 datacenters left so it's all good. gotta keep those ad dollars flowing!"
|
# ¿ May 18, 2016 17:46 |
|
i remember reading this a while ago. the key thing to take away from this is that the most common source of errors is fleshy imperfect meatbags and we need to get rid of them asap the dbas mentioned in this thread seem to corroborate that
|
# ¿ May 18, 2016 19:18 |
|
i can see where they're coming from. i've seen db schemas with so much poo poo in them i wouldn't be surprised if they'd been created 6000 years ago lol at them being intelligently designed though
|
# ¿ May 18, 2016 23:27 |
|
no. i don't need more poo poo on my reading list
|
# ¿ May 20, 2016 20:10 |
|
time for me to shill for "Readings in Database Systems" which is pretty much a collection of classic or otherwise interesting db papers. it's pretty good
|
# ¿ May 20, 2016 20:43 |
|
yeah someone should pass a law that forces the universe to make all values depend on at most two variables. would solve this whole thing right quick
|
# ¿ May 22, 2016 17:32 |
|
OCC is stupid crap for naive idiots who have yet to feel the full wrath of murphy's law. "no, really, we basically don't have any contention, it's gonna be fine guys!" bullshit you need to lock everything down, make sure everything happens exactly when it should and enforce a new order upon your database with the lock manager acting as a ruthless fascist regime. if two transactions need locks owned by one another, you take one of them out back and shoot it. that's why it's called a deadlock, because one of your transactions will die
|
# ¿ May 26, 2016 16:44 |
|
|
# ¿ Apr 28, 2024 21:13 |
|
Cocoa Crispies posted:you have two cluster members, "piper" and "zedo" that's not really a transaction and 2 phase locking or optimistic concurrency control issue, though. you're asking a question about the system as a whole (which transaction is seen first by the system), but don't provide a mechanism to establish a total order for events in the system as a whole. so of course there's no valid answer
|
# ¿ May 27, 2016 16:46 |