[chef] Re: Re: Re: mysql-server fails on first restart within install

[Adam Jacob < >]

On Sat, Aug 8, 2009 at 2:25 AM, Matthew 
Todd<
 >
 wrote:
> Running the following command, which has the same chpst memory limit as the
> chef-client runit service, MySQL server FAILS to create the initial InnoDB
> tables:
>
>        chpst -m 262144000 sh -c '/usr/bin/env DEBIAN_FRONTEND=noninteractive
> apt-get -q -y install mysql-server'
>
> If I remove the "-m 262144000", however, it SUCCEEDS.
>
> Strangely, though, doubling the memory limit to 524288000 (512M), 1048576000
> (1G), and 2097152000 (2G) doesn't help: MySQL server still FAILS to restart.
>
> (And when I use "chpst -d ..." rather than "-m" (just limiting the data
> segment), it SUCCEEDS.)

Wow, nice troubleshooting.

You can safely remove the memory limit in the chef-client run file,
and this might be a fine reason to do so.  One interesting thing here
is that we might be having our children inherit the limits set by
chpst, which would be Bad (tm) for something like Chef.

> So, for now, I've draconically removed the chpst -m memory limit from
> chef-client's runit script -- and mysql-server installs just fine.
> Long-term, though, I feel a little uneasy, since I'm sure that limit was
> there for a reason.

I think it's been there mainly as a form of garlic to ward off demons
from our collective past (ie: rapid crazy memory bloating of our long
running configuration management processes.)  We haven't had many
reports of this happening with Chef, though, so I think it should be
safe to remove it.

0.7.6 also introduces the ability to use the splay setting without an
interval, which means you could just use cron to get the same
functionality as the long-running chef-client, if that sort of thing
suits you. (It does make things harder to turn off, which is why I
don't take that path personally.)

Best,
Adam

-- 
Opscode, Inc.
Adam Jacob, CTO
T: (206) 508-7449 E: