launchpad-dev team mailing list archive

[John Arbash Meinel <john@xxxxxxxxxxxxxxxxx>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 1/30/2011 4:27 PM, Robert Collins wrote:
> On Mon, Jan 31, 2011 at 11:14 AM, Michael Hudson-Doyle
> <michael.hudson@xxxxxxxxxxxxx> wrote:
>> If there's anything specific I can do to help with getting the
>> history-db stuff deployed, please ask.  I'm pretty convinced it's an
>> improvement in every way we actually care about.
> 
> So, I'm fiddling with the branches now - having finally gotten on top
> of the bugs :)
> 
> What we need is a block of engineer time:
>  - finish stabilising loggerhead (get to a stable baseline)
>    specifically bug 708961 and bug 701329 combine to make it very hard
> to detect regressions here
>  - check we're not running into cache locking issues etc (which
> history db could make explode)
>  - configure for historydb
>  - land historydb in the lp loggerhead branch
>  - merge to devel
>  - qa aggressively (may require qa changes to get multiple processes)
>  - get deployed and watch like a hawk for a few days.
> 
> -Rob

I'm certainly willing to do some of this.

Cache locking issues- potentially gets a lot better. If 2 loggerhead
processes want to access the same branch at the same time (at a new tip
revision), they'll end up spending 6+s to rebuild the cache. The
history-db code can spend ~100ms doing the update instead of grabbing
the whole history.

Sharing the caches can introduce the same effect across branches. You
have a higher chance of 2 processes wanting the same resource, but the
time spent is much lower.

Ultimately, I think the primary concern is how to get serialized access
to write access. Which sqlite probably gives us by blocking BEGIN if
someone else is running it. We can even look at removing the in-process
exclusive lock, since many of the threads could actually be serving
different branches, and thus different caches.



There is one other thing that I'm a bit worried about. I don't know what
changed, but I've started getting bounces from
launchpad-bugs-owner@xxxxxxxxxxxxxxxxxxx whenever I try to post code for
review. And I noticed that once I started getting that, I stopped
getting reviews.

Did you change the default reviewer? In changing that, did you not
actually allow access for those reviews to be sent to the new location?

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk1MGbQACgkQJdeBCYSNAAMlkwCfSON+nIXupttjFshdtuVbLgGv
rk0AnRbldxG9DOJSLnjrdMoKccg8aOjU
=pFli
-----END PGP SIGNATURE-----