FreeBSD support being added to GlusterFS

[Jordan Hubbard]

Thanks!  Just to make sure I understand for our own ports repo in FreeNAS-land:

1. You have updated the tarball but Baptiste is still maintaining the port, correct?

If Baptiste would like someone else to take it over, I can probably coax one of our resident committers to do it as well.  We frankly have not been able to do any testing of glusterfs in FreeNAS yet since glusterd just catches a signal 10 whenever any glusterfs operation is performed.  Since this is also really early BETA software, I figured the FreeBSD port was still a bit green and I'd simply wait for it to get a bit more mature.  Now that you say you’ve gotten a regression suite up and running, I’m guessing the time to really push on it in FreeNAS 9.3-ALPHA is probably soon?

2. That said, it also sounds like updating the glusterfs port is going to have to wait for your cmockery2 port to get accepted and committed to the tree, since otherwise it won’t build - is that a correct assessment?

Cheers,

- Jordan

On Aug 11, 2014, at 1:25 AM, Harshavardhana <harsha at harshavardhana.net> wrote:

> http://download.gluster.org/pub/gluster/experimental/glusterfs-freebsd_20140811.tar.bz2
> - updated tar with
> 
> - libexecinfo part of glusterfs 'contrib'
> - Gluster configuration changed from "/var/lib" to "/var/db" by
> default for non-linux platforms.
> - First attempt to get regression tests ported to FreeBSD
> - Currently the build is dependent on 'cmockery2' for which i
> submitted a new port request here -
> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=192420 (not merged
> yet)
> 
> Let me know how the testing goes.
> 
> BTW as an off topic but related to GlusterFS regression tests - i have
> been trying to get 'pidof' command really working on FreeBSD but seems
> like its really un-usable and doesn't behave the same way as the
> sysvinit `pidof` does.
> 
> 
> On Sun, Jul 6, 2014 at 4:09 PM, Baptiste Daroussin <bapt at freebsd.org> wrote:
>> On Sun, Jul 06, 2014 at 12:13:15PM -0700, Harshavardhana wrote:
>>>> 
>>>> I can make the /usr/local/var/log/glusterfs directory and it gets much further.   That said, is there some special configure flags we should be passing in our version of the port to properly stuff glusterfs into /var instead?  Your email tends to imply that we should be passing —localstatedir, which we can certainly do no problem, I’m just wondering if that’s your long-term plan.  Again, this is our port:  https://github.com/freenas/ports/tree/freenas/9-stable/sysutils/glusterfs
>>>> 
>>>> The fundamental issue with /usr/local is, again, that /usr/local is read-only on FreeNAS.  If there are configuration files that glusterfs expects to be modifiable, they can’t live anywhere in /usr/local, nor of course can any temporary files or log files.  We have made special provisions for /etc and /var such that those can be modified, so we basically just need to compile gluster as a “system service” and put it in the system directories (e.g. prefix is /, not /usr/local).
>>>> 
>>> 
>>> Ah now i get it - "/usr/local" is not a requirement for "GlusterFS" it
>>> is a baggage of using "autotools" when during ./configure if you do
>>> not specify --prefix - so for a standard installation under RPM it is
>>> usually the following flags are used
>>> 
>>> # ./configure --prefix=/usr --sysconfdir=/etc --localstatedir=/var
>>> --libdir=/usr/lib64
>>> 
>>> Since FreeBSD doesn't need "/usr/lib64" you could just use for packages
>>> 
>>> # ./configure --prefix=/usr --sysconfdir=/etc --localstatedir=/var
>>> 
>> 
>> Here is an updated version of my port
>> http://people.freebsd.org/~bapt/glusterfs.diff
>> 
>> This time it passes poudriere (for those not aware of poudriere it is for
>> FreeBSD a bit like what mock is for fedora but on steroid :))
>> 
>> What is new in there: dependency on bison that I missed the first time, a
>> dependency on libexecinfo (on non FreeBSD 10) and a build dependency on git
>> other build-aux/pkg-version is not happily catching the version
>> 
>> Tested on FreeBSD 10
>> 
>> regards,
>> Bapt
> 
> 
> 
> -- 
> Religious confuse piety with mere ritual, the virtuous confuse
> regulation with outcomes