On Wed, Sep 15, 2010 at 8:26 AM, John Arbash Meinel <[email protected]> wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > So I've finally gotten the end-to-end code working, such that 'ssh lp > bzr serve ...' can actually use a daemon which preloads some python > libraries, forks for each request, and doesn't re-exec. > > lp:~jameinel/launchpad/lp-service > > > I think there is still a fair number of cleanup passes necessary, but I > think for all of that I'm going to need a bit of help. Stuff like "where > should I be putting the log files", etc. > > As such, what is the best way to progress forward? Put it up as a merge > request? Get someone to pair with me for a little while to have a quick > turn around for all those little questions?
I suggest: - get it reviewed and then either: - guard it with a featureflag so its inactive, and land it disabled to be iterated on - keep tuning it until everything seems honky dory and then land with shock and awe. > Times for 'echo hello | ssh localhost -p 5022 bzr serve' > > 2.5s baseline > 0.8s without twisted conch and 'lp-serve' > 0.6s 'ssh localhost /bin/false' (openssh) > 0.25s with forking -nice-. > So at this point, connecting to the 'lp' service is faster than running > against openssh and spawning bzr directly. And, in general, it looks to > shave most of the connection overhead. (The 250ms seems to be mostly > fork() time, as lsprof only sees about 50ms in the client.) This is awesome stuff and I'm so very very glad you've been working on it. -Rob _______________________________________________ Mailing list: https://launchpad.net/~launchpad-dev Post to : [email protected] Unsubscribe : https://launchpad.net/~launchpad-dev More help : https://help.launchpad.net/ListHelp
