On Sun, Feb 19, 2012 at 7:54 AM, Vadim Tkachenko <[email protected]> wrote:
> Hello,
>
> I tried to play with drizzle, and as my experience is mostly
> unpleasant, as my goal left non-achieved.
> The goal was to run the similar benchmark as described in
> http://www.mysqlperformanceblog.com/2012/02/18/mariadb-5-3-4-benchmarks/
>
> I decided to share it with you, in case you want an opinion from a third
> party.
I saw the bugs you filed, thank you for that.
>
> I should say that I used RPM
> drizzle7-2012.01.30-1.el6.x86_64.rpm.tar.gz . Probably some issues
> resolved in more recent RC.
I think it's very unfortunate that building RPMs and DEBs was
neglected for so long. It really would have been more useful if you
could have tried the RC. Anyway, I hope that it's just a matter of
days before first packages are published for 7.1 series.
In some ways I'm even happy you gave up on the 7 release, since it is
old by Drizzle standards. Would be much more interesting to have
results for Drizzle 7.1. (I'm not saying I know it to be better, just
that the results would be more useful and comparable.)
Out of the bugs you filed, it's possible some of them have been fixed,
but I think I have seen the --no-defaults not working also with the
current trunk. I was doing something else so I didn't chase it down,
but that one seems familiar.
Basically, my experience is that Drizzle isn't generally very
unstable, but it's also not very tested in various (often common) end
user scenarios. Options being undocumented, or not working as
documented (or at all) is a common example. Getting bug reports from
users who weren't involved in developing the code themselves is the
best cure. Now the real test is whether the bugs are fixed or not.
(Last RC closed 20-30 bugs, so I'm optimistic.)
Otoh the cond_wait behavior you reported seems more like a bad bug and
will be interesting to see what the cause of that was, and how quickly
someone rushes to fix it.
> ( mysql -e "select avg(id) from sbtest$i FORCE KEY (PRIMARY); " sbtest
> ) > $OUTDIR/warmup.${i}.out 2>&1 &
>
> This is 16 connections running in parallel.
>
> This process never finished in drizzle,
> as it seems it was serialized somewhere and I saw only 1 thread running.
Seems like you didn't report this one anymore, so I did:
https://bugs.launchpad.net/drizzle/+bug/936409
henrik
--
[email protected]
+358-40-8211286 skype: henrik.ingo irc: hingo
www.openlife.cc
My LinkedIn profile: http://www.linkedin.com/profile/view?id=9522559
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
