sks-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Sks-devel] Min. Requirement for SKS Version in the Pool


From: Jeffrey Johnson
Subject: Re: [Sks-devel] Min. Requirement for SKS Version in the Pool
Date: Mon, 02 Jul 2012 11:01:46 -0400

On Jul 2, 2012, at 9:54 AM, Stefano Rivera <address@hidden> wrote:

> 
> * And the related "unable to allocate memory for mutex; resize
>  mutex region" issue has also been reported in Ubuntu.
>  https://bugs.launchpad.net/ubuntu/+source/sks/+bug/937931
>  I didn't touch this, as the solution seems to be to change the way we
>  upgrade the DB, and I'd like to not deviate from Debian there.
> 

This is actually the single most reported BDB problem and
should be addressed for "drop-in" SKS usage cases somehow.

The hard problem here is that the values needed
are extrinsic (i.e. dependent on #of keys),
interrelated with other parameters (half the pagesize
and you will need 2x the locks), as well as infrastructure
related (running on 512MB instead of 16GB usually needs
drastically different tuning).

But its not too hard to get the tuning mostly right
with a heuristic by:
        1) make some assumption like 25% of available memory can be used
        2) use pagesize/no. of locks together
        3) examine real world usage cases and apply a 2x
        engineering threshhold.

The heuristic MUST be run-time, not build-time, because
of "25% of available memory" scaling.

73 de Jeff




reply via email to

[Prev in Thread] Current Thread [Next in Thread]