[HCoop-Discuss] Reorganizing, people-wise and tech-wise

[John T. Settino]

> On Fri, Jun 26, 2009 at 02:25:41PM -0400, John T. Settino wrote:
>>
>> > My opinion (because it looks like people want opinions here):
>> >
>> > I was/am definitely one of the hcoop members put off by relearning AFS
>> and
>> > domtool stuff when I wanna add something to my account.  The main
>> problem
>> > with AFS that I have isn't that it's networked (which is a great idea)
>> but
>> > that I can't use the usual unix permission commands to manipulate my
>> > files.
>> >
>>
>> Same here. No qualms with a networked filesystem (we use NFS at work),
>> but
>> the ACLs were definitely an unwelcome change (had to modify portions of
>> my
>> website so it would continue working). For a coop made up of "power
>> users"
>> who I assume are very familiar with UNIX permissions, moving to an
>> architecture that we don't have but one person to admin (says a lot
>> about
>> the technology vs our userbase, right there!), and forcing the rest of
>> the
>> userbase to learn a new way of doing things, was pretty stupid IMO. Same
>> goes for domtool v1 -> domtool2. No effort to help the users convert
>> their
>> existing configs to new ones, just a "learn it or tough crap" brush off.
>
> I agree, parsing the output of fs listacl correctly is horrible and the
> restriction that ACLs can only be set on directories is a disadvantage
> too.
>
>> I believe an ideal architecture would be as follows:
>>
>> 2 identical NAS machines for data storage w/ NFS
>
> So you volunteer to set up and maintain NFS with LDAP? That would be
> great.

I WISH I knew enough to set up and maintain this. I am very slowly
learning LDAP and such because that is what we use at work for our
authentication system (OpenLDAP to be specific), but I do not know it
enough to administrate it myself.
>
>> 2 identical beefy machines, each housing virtualized servers for
>> services
>> VMs would be something like
>> mail
>> web
>> db
>> shell
>> ldap (for auth)
>> et. al.
>
> Do we really need VMs? I know that virtualisation is hyped these days,
> but they should be used with care. Just because we have several VMs,
> we don't avoid a single point of failure.
This is true, but do better solutions exist for dedicated a specific
amount of CPU and memory to specific applications? Having a dedicated VM
for each service can prevent one haywire service from taking down an
entire infrastructure, as what happened when AFS borked on delueze the
other day. If for example, email was fscking up, you could reboot the mail
VM and not have to restart each and every service at the same time because
you rebooted the machine they were all housed on.
>
> Maybe we should make a volunteer list that lists persons, skill and
> desired responsibilities.
Agreed.
>
>> Duplicates of everything for failover capabilities, obviously.
>
> The wiki lists rsync.net as a remote backup provider, that sounds like a
> good idea.
>
Does anyone have an accurate listing of just how much data the coop is
using? I'm not sure. This would definitely indicate the size of NAS or
storage devices we would need to purchase for any RAID arrays.

>> - J
>
> Regards,
> Matthias-Christian
>
> _______________________________________________
> HCoop-Discuss mailing list
> HCoop-Discuss at lists.hcoop.net
> https://lists.hcoop.net/listinfo/hcoop-discuss
>
>


-- 
John T. Settino
Web Developer / Technologist for Hire
http://www.johnsettino.com/

* Please consider the environment before printing this email. :DGC