Re: [new] databases/influxdb

On Fri, Dec 29, 2017 at 07:11:55PM +0100, Landry Breuil wrote:
> On Fri, Dec 29, 2017 at 07:45:41AM +0100, Landry Breuil wrote:
> > On Thu, Dec 28, 2017 at 11:34:31PM +0100, Antoine Jacoutot wrote:
> > > On Thu, Dec 28, 2017 at 09:14:20PM +0000, Landry Breuil wrote:
> > > > On Wed, Dec 27, 2017 at 05:43:39PM +0000, Stuart Henderson wrote:
> > > > > On 2017/12/25 07:34, Landry Breuil wrote:
> > > > > > On Sun, Dec 24, 2017 at 08:27:11PM +0000, Stuart Henderson wrote:
> > > > > > > On 2017/12/24 10:35, Landry Breuil wrote:
> > > > > > > > Hi,
> > > > > > > >
> > > > > > > > here's a port for influxdb 1.4.2, a time series database to store
> > > > > > > > metrics/events/analytics. I had to fight a bit with go.port.mk, and hack
> > > > > > > > my way around bundling the dependencies (which are vendored in the
> > > > > > > > selfhosted tarball) but it builds, installs and starts fine.
> > > > > > > >
> > > > > > > > I haven't done any real testing yet (read: send metrics to it and try to
> > > > > > > > read them) but it can be used as a target for collectd network plugin,
> > > > > > > > ie replacing old-school rrd storage (no need for rrdcached anymore!),
> > > > > > > > and facette is supposed to be able to read metrics from it. Of course,
> > > > > > > > the next step is to port grafana..
> > > > > > > >
> > > > > > > > Testing & feedback welcome!
> > > > > > > >
> > > > > > > > Landry
> > > > > > >
> > > > > > > GH_* shouldn't be set alongside MASTER_SITES .. Probably should just be
> > > > > > > the usual DISTNAME and/or PKGNAME ..
> > > > > >
> > > > > > Well, for now it's more a convenience so that i can fetch the original
> > > > > > source from github (without the vendored dependencies) - this way i only
> > > > > > need to comment MASTER_SITES/EXTRACT_SUFX and run make tarball.
> > > > > > Maybe a comment would help ? Or commented out values ?
> > > > > >
> > > > >
> > > > > Probably better if that fetching is handled in 'make tarball' I think,
> > > > > it's even more confusing to temporarily use ports distfile fetching as an
> > > > > intermediate step to producing the actual ports distfile..
> > > >
> > > > Well, i tried to remove the GH_* stuff and put all the gory bits of
> > > > tarball repacking in this make tarball target, but that becomes *very*
> > > > ugly. Having GH_ACCOUNT/GH_PROJECT set triggers lots of magic in
> > > > go.port.mk (whether it's a good idea or not is another topic) that the
> > > > other targets of the port sort of rely on (setting WRKSRC correctly,
> > > > among others.. ALL_TARGET, that weird MODGO_SETUP_WORKSPACE thing) so i
> > > > end up reinventing some bits.
> > > >
> > > > Here's what the a diff between two variations of the Makefile would look
> > > > like. And i have to admit i find this ugly :)
> > >
> > > Why not ask upstream to add vendor support?
> >
> > There's https://github.com/influxdata/influxdb/issues/7696 open since a
> > year, so i wouldnt count on this happening in the short term.
> > Distributing software via a source tarball is dead anyway, everyone uses
> > docker nowadays, right ?
>
> So here's a new version incorporating the above changes, with a complete
> make tarball target that i'd probably be the only one to use.
> Repackages/builds fine here, and i'm running it on -current and 6.2 on
> my home router. If we want grafana in the portstree, would be good to
> have influxdb first, so looking for feedback and okays :)
>
> Landry
> >

Works fine for me. I tested it on -current amd64 with collectd sending
metrics to influxdb.

It's a pity that influxdata does not include the dependencies in their src
tarball. In their build.py the do the same ting with gpm as you do int the
tarball target.

Maybe we need something like GH_TUPLE in FreeBSD for go deps.

I like the simpler tarball target in Makefile.github but I understand
sthen@'s arguments against it.

ok remi@ for the version without GH_*