launchpad-dev team mailing list archive

[Jeroen Vermeulen <jtv@xxxxxxxxxxxxx>]

On 2011-07-15 00:27, Robert Collins wrote:
> As prep for https://dev.launchpad.net/LEP/FastDowntime the db schema

> I appreciate that between now and the deployment side of this going
> live we're going to be doing a little more work around each schema
> patch with no immediate payoff - but it means we'll be able to go
> straight into the fastdowntime process once its ready.

Thank you so much for pushing this through!

There _is_ an immediate payoff: your db patch gets deployed much faster. 
 We've been aching for this kind of flexibility.

I ran some numbers a few years back when I proposed weekly rollouts of 
devel plus "live" DB patches.  For a medium-sized change, live patches 
took more work but the worst-case time to useful deployment was about 
the same as the best-case time for our current scheme.  Typical cases 
were massively better.  And that was with a limitation of one phase per 
week; we're getting something much faster.

We'll learn how to write incremental patches.  Meeting the 15s goal on a 
cold database will be hard though, since we can't easily reproduce 
results or try out tweaks.  It might help if we could break a slave out 
of replication to time a DB patch against a realistic cache.


Jeroen