[SAC] Backup Implementation for OSGeo VMs

Committees SAC mailing list
1

Lance & Team,

Right now OSGeo is running our own Bacula on one of our VMs to backup content from our other VMs. We noticed that OSUOSL also runs Bacula on a larger scale and was wondering if we could levarage that in some way. We had been discussing buying a cheap 1U node just for backup but could just as easily contribute disks or funds to your setup if it achieves the same results.

We would like to be able to tweak or configure as needed but expect that dumping our current config to your bacula would suffice for now.

What route does OSUOSL suggest?

Thanks,
Alex
OSGeo Systems Administration Committee

2

On Tue, Feb 28, 2012 at 3:05 PM, tech@wildintellect.com via RT <support@osuosl.org> wrote:

Right now OSGeo is running our own Bacula on one of our VMs to backup
content from our other VMs. We noticed that OSUOSL also runs Bacula on a
larger scale and was wondering if we could levarage that in some way. We
had been discussing buying a cheap 1U node just for backup but could
just as easily contribute disks or funds to your setup if it achieves
the same results.

We would like to be able to tweak or configure as needed but expect that
dumping our current config to your bacula would suffice for now.

What route does OSUOSL suggest?

I would prefer integrating it into our bacula and we could come up with a cost estimate for you to contribute into. How much data are you backing up? What kind of a rotation are you using/doing? How many clients need to be backed up? We also sync our bacula data off-site daily.

–
Lance Albertson
Systems Administrator / Architect

3

Hi Lance,

On Tue, Feb 28, 2012 at 11:07:53PM +0000, Lance Albertson via RT wrote:

I would prefer integrating it into our bacula and we could come up with a
cost estimate for you to contribute into. How much data are you backing up?
What kind of a rotation are you using/doing? How many clients need to be
backed up? We also sync our bacula data off-site daily.

The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
even though we currently don't use all of it, I think it's a good idea
being prepared to do so.

My favourite is to do one full backup per week, differentials every day
and to keep a full backup at least until the successor has finished, at
best until shortly before n+2 gets written to the media.
I was informed that we don't require retaining months of backup
history. Thus the space requirements would account to approx.
4.5 - 5 TByte (two full backups plus one busy week of differentials).

Currently we're backing up approx. 10 clients, I'd expect the number to
grow slightly.

Best regards,
  Martin.
--
Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------

4

Hi Lance,

On Tue, Feb 28, 2012 at 11:07:53PM +0000, Lance Albertson via RT wrote:

I would prefer integrating it into our bacula and we could come up with a
cost estimate for you to contribute into. How much data are you backing up?
What kind of a rotation are you using/doing? How many clients need to be
backed up? We also sync our bacula data off-site daily.

The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
even though we currently don't use all of it, I think it's a good idea
being prepared to do so.

My favourite is to do one full backup per week, differentials every day
and to keep a full backup at least until the successor has finished, at
best until shortly before n+2 gets written to the media.
I was informed that we don't require retaining months of backup
history. Thus the space requirements would account to approx.
4.5 - 5 TByte (two full backups plus one busy week of differentials).

Currently we're backing up approx. 10 clients, I'd expect the number to
grow slightly.

Best regards,
  Martin.
--
Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------

5

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org> wrote:

The combined net disk space on the OSGeo VM’s is approx. 2 TByte and,
even though we currently don’t use all of it, I think it’s a good idea
being prepared to do so.

That’s going to take a nice chunk of space on our server but its

6

Lance,

Attached is a current copy of our configuration. I'll post other answer
more publicly on another email.

Thanks,
Alex

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org>wrote:

The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
even though we currently don't use all of it, I think it's a good idea
being prepared to do so.

That's going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the costs of
having us do the backups? I can give you an estimate for that if you'd like.

My favourite is to do one full backup per week, differentials every day
and to keep a full backup at least until the successor has finished, at
best until shortly before n+2 gets written to the media.
I was informed that we don't require retaining months of backup
history. Thus the space requirements would account to approx.
4.5 - 5 TByte (two full backups plus one busy week of differentials).

We tend to retain more backups ourselves but we can have a specific
schedule for your backups that meet your needs. We currently use SSL
enabled connections so we'll have to give you some keys to get it to work
with our setup. Do you have any special FileSet filters you use so you
don't backup stuff like logs?

Currently we're backing up approx. 10 clients, I'd expect the number to
grow slightly.

*nods* Currently restores have to be initiated by us but we are working
towards creating access to work around that. Will this be a problem?

Thanks-

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org
<mailto:support@osuosl.org>> wrote:

    The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
    even though we currently don't use all of it, I think it's a good idea
    being prepared to do so.

That's going to take a nice chunk of space on our server but its

bacula-dir.conf (11.8 KB)

7

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org>wrote:

The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
even though we currently don't use all of it, I think it's a good idea
being prepared to do so.

That's going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the costs of
having us do the backups? I can give you an estimate for that if you'd like.

Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever units
make sense. The alternative we've been considering is buying a new server.

My favourite is to do one full backup per week, differentials every day
and to keep a full backup at least until the successor has finished, at
best until shortly before n+2 gets written to the media.
I was informed that we don't require retaining months of backup
history. Thus the space requirements would account to approx.
4.5 - 5 TByte (two full backups plus one busy week of differentials).

We tend to retain more backups ourselves but we can have a specific
schedule for your backups that meet your needs. We currently use SSL
enabled connections so we'll have to give you some keys to get it to work
with our setup. Do you have any special FileSet filters you use so you
don't backup stuff like logs?

A copy of our current bacula director config has been sent to Lance. In
some cases (e.g. Apache) we actually want to keep some logs. We are
still undergoing some refinement to make the whole thing more efficient.

I assume you can take care of the keys, since you have access to all the
guests.

Currently we're backing up approx. 10 clients, I'd expect the number to
grow slightly.

*nods* Currently restores have to be initiated by us but we are working
towards creating access to work around that. Will this be a problem?

Thanks-

Generally no, we hope to never have a need for such a restore and in
such an event we'd likely be filing a ticket with you anyways because
something went wrong with the hardware or VMs. Though access to extract
some files from the backups would be nice long term.

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org
<mailto:support@osuosl.org>> wrote:

    The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
    even though we currently don't use all of it, I think it's a good idea
    being prepared to do so.

That's going to take a nice chunk of space on our server but its

We're more than happy to buy disks if that's the best solution.

Thanks,
Alex

8

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org>wrote:

The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
even though we currently don't use all of it, I think it's a good idea
being prepared to do so.

That's going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the costs of
having us do the backups? I can give you an estimate for that if you'd like.

Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever units
make sense. The alternative we've been considering is buying a new server.

My favourite is to do one full backup per week, differentials every day
and to keep a full backup at least until the successor has finished, at
best until shortly before n+2 gets written to the media.
I was informed that we don't require retaining months of backup
history. Thus the space requirements would account to approx.
4.5 - 5 TByte (two full backups plus one busy week of differentials).

We tend to retain more backups ourselves but we can have a specific
schedule for your backups that meet your needs. We currently use SSL
enabled connections so we'll have to give you some keys to get it to work
with our setup. Do you have any special FileSet filters you use so you
don't backup stuff like logs?

A copy of our current bacula director config has been sent to Lance. In
some cases (e.g. Apache) we actually want to keep some logs. We are
still undergoing some refinement to make the whole thing more efficient.

I assume you can take care of the keys, since you have access to all the
guests.

Currently we're backing up approx. 10 clients, I'd expect the number to
grow slightly.

*nods* Currently restores have to be initiated by us but we are working
towards creating access to work around that. Will this be a problem?

Thanks-

Generally no, we hope to never have a need for such a restore and in
such an event we'd likely be filing a ticket with you anyways because
something went wrong with the hardware or VMs. Though access to extract
some files from the backups would be nice long term.

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org
<mailto:support@osuosl.org>> wrote:

    The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
    even though we currently don't use all of it, I think it's a good idea
    being prepared to do so.

That's going to take a nice chunk of space on our server but its

We're more than happy to buy disks if that's the best solution.

Thanks,
Alex

9

On Thu, Mar 1, 2012 at 4:49 PM, tech@wildintellect.com via RT <support@osuosl.org> wrote:

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org>wrote:

The combined net disk space on the OSGeo VM’s is approx. 2 TByte and,
even though we currently don’t use all of it, I think it’s a good idea
being prepared to do so.

That’s going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the costs of
having us do the backups? I can give you an estimate for that if you’d like.

Yes, we’d like an estimate per GB, or 100 GB or 1 TB. Whichever units
make sense. The alternative we’ve been considering is buying a new server.

So I just re-did all our “rates” recently to figure out costs and it boils down to the following;

10

From Lance's response I conclude that we're going to roll our own

backup, simply because their annual fee for backing up the OSGeo VM's
by OSL is about twice our single budget for buying a dedicated backup
machine.

Anyhow I've never understood how OSGeo got into financial obligations
with respect to the hosting at OSL.
When the new servers were purchased, everbody was talking about free
hosting at OSL in the sense of "no matter whatever would happen to
OSGeo, the server infrastructure won't be affected and will continue".
Now recently I realized that OSGeo is paying an annual fee of $3.5k for
hosting the two servers at OSL - which is pretty contradictory to what
has been communicated before.
Thus if OSGeo would buy another machine for doing the backup, I
understand that OSGeo would have to pay another annual fee for hosting
the third machine.

Don't get me wrong, I don't object against the figure on the price tag,
instead I'm aware that the fees at OSL are hard to beat. Anyhow I
wonder wether the additional costs for hosting the third machine had
been taken into account ....

Cheers,
  Martin.
--
Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------

11

On 03/02/2012 02:45 AM, Martin Spott wrote:

From Lance's response I conclude that we're going to roll our own

backup, simply because their annual fee for backing up the OSGeo VM's
by OSL is about twice our single budget for buying a dedicated backup
machine.

Anyhow I've never understood how OSGeo got into financial obligations
with respect to the hosting at OSL.
When the new servers were purchased, everbody was talking about free
hosting at OSL in the sense of "no matter whatever would happen to
OSGeo, the server infrastructure won't be affected and will continue".
Now recently I realized that OSGeo is paying an annual fee of $3.5k for
hosting the two servers at OSL - which is pretty contradictory to what
has been communicated before.

We are not under obligation to do so as far as I know. We offered a one
time contribution for their services and assistance to help ensure
future hosting.

Thus if OSGeo would buy another machine for doing the backup, I
understand that OSGeo would have to pay another annual fee for hosting
the third machine.

Don't get me wrong, I don't object against the figure on the price tag,
instead I'm aware that the fees at OSL are hard to beat. Anyhow I
wonder wether the additional costs for hosting the third machine had
been taken into account ....

Cheers,
  Martin.

I agree the original quote is a little high considering out
alternatives. So some questions, do we really need 5 TB? or would 2.5TB
do for now? Anyone have a counter-proposal we could offer?

The 2 enticing factors about OSUOSL is the buying in to staff time and
offsite backup. 2 things that are not accounted for in buying our own
server.

Now of course the alternate idea is to get a 1U 4 drive x 2/3 TB SATA
drives and run it ourselves. Estimated cost is $2-2.5k
Which we could host at OSUOSL or as discussed before put elsewhere so
the backup is off-site to begin with (Bacula does have a bandwidth
throttle option, and I don't think we move that much data per night to
make it a huge issue, though moving a full 1 TB once a week may not work)

Other opinions, please add in so we can come up with an approach to
bring to the board, and refine our negotiations.

Thanks,
Alex

12

On Fri, Mar 02, 2012 at 08:59:33AM -0800, Alex Mandel wrote:

I agree the original quote is a little high considering out
alternatives. So some questions, do we really need 5 TB? or would 2.5TB
do for now?

This doesn't make a difference for us. They're asking $50 per machine
per month, thus by having so many different VM's, the total backup
space isn't the primary concern.

Now of course the alternate idea is to get a 1U 4 drive x 2/3 TB SATA
drives and run it ourselves. Estimated cost is $2-2.5k
Which we could host at OSUOSL or as discussed before put elsewhere so
the backup is off-site to begin with (Bacula does have a bandwidth
throttle option, and I don't think we move that much data per night to
make it a huge issue, though moving a full 1 TB once a week may not work)

For doing serious backup I'd consider pulling a full dump on a regular
schedule a requirement because just chaining several weeks or months of
incrementals together on top of a terribly old full dump is pretty
risky.

Cheers,
  Martin.
--
Unix _IS_ user friendly - it's just selective about who its friends are !
--------------------------------------------------------------------------

13

On Fri, Mar 2, 2012 at 8:59 AM, Alex Mandel <tech_dev@wildintellect.com> wrote:

On 03/02/2012 02:45 AM, Martin Spott wrote:

Anyhow I've never understood how OSGeo got into financial obligations
with respect to the hosting at OSL.
When the new servers were purchased, everbody was talking about free
hosting at OSL in the sense of "no matter whatever would happen to
OSGeo, the server infrastructure won't be affected and will continue".
Now recently I realized that OSGeo is paying an annual fee of $3.5k for
hosting the two servers at OSL - which is pretty contradictory to what
has been communicated before.

We are not under obligation to do so as far as I know. We offered a one
time contribution for their services and assistance to help ensure
future hosting.

Alex / Martin,

Note that it was my intention that we would annually provide a contribution
to OSL as a way of showing some appreciation for the services they
provide even if it doesn't actually cover what we cost them to serve.
However, it is not an obligation.

I will note the amounts of money Lance mentioned for supporting backup
seem high to me, and I'm not too keen on getting into a situation with
high locked in prices for backup services.

Best regards,
--
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up | Frank Warmerdam, warmerdam@pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush | Geospatial Software Developer

14

On March 2, 2012 at 2:32 PM Frank Warmerdam <warmerdam@pobox.com> wrote:

On Fri, Mar 2, 2012 at 8:59 AM, Alex Mandel <tech_dev@wildintellect.com>

wrote:

> On 03/02/2012 02:45 AM, Martin Spott wrote:
>> Anyhow I've never understood how OSGeo got into financial obligations
>> with respect to the hosting at OSL.
>> When the new servers were purchased, everbody was talking about free
>> hosting at OSL in the sense of "no matter whatever would happen to
>> OSGeo, the server infrastructure won't be affected and will continue".
>> Now recently I realized that OSGeo is paying an annual fee of $3.5k

for

>> hosting the two servers at OSL - which is pretty contradictory to what
>> has been communicated before.
>
> We are not under obligation to do so as far as I know. We offered a one
> time contribution for their services and assistance to help ensure
> future hosting.

Alex / Martin,

Note that it was my intention that we would annually provide a

contribution

to OSL as a way of showing some appreciation for the services they
provide even if it doesn't actually cover what we cost them to serve.
However, it is not an obligation.

I will note the amounts of money Lance mentioned for supporting backup
seem high to me, and I'm not too keen on getting into a situation with
high locked in prices for backup services.

Best regards,
--

I agree that at that price it's not a good idea, but I also think there
might be some room here for negotiation.
Checking Amazon S3 prices, 5 TB would run us $550 per month at the regular
rate. So the price isn't astronomical.
The odd point for me, is that we have multiple VMs based on their advice,
technically we only have 2 machines, if you count clients that way we are
talking 1 TB of space at $100/month, if we need 2 TB then maybe we could
negotiate for say $150-200 a month (1800-2400/yr) which seems much more
reasonable.

We should also realize we don't backup everything this way. The download
server has long operated on make an rsync of it. Many of the old foss4g
websites and uploads contained are archival in nature, not changing. So I
think we should calculate out what's static and what's dynamic.

Basically a full dump of the downloads folder every week seems silly, but a
full dump of various postgres/mysql databases running sites and compressed
on dump seems good, but that won't take the same space it takes on the
running server. So I think estimating disk space based on a full 1:1 of the
complete amount of disks we have is quite an overestimation of our needs.

Just want to make sure we do a good comparison before deciding and factor
things in like "cost" of maintaining the backups.

Thanks,
Alex

15

We're you still want to do this? We can always work out a better deal if
the cost is too high.

Thanks
On Mar 1, 2012 5:07 PM, "Lance Albertson" <ramereth@osuosl.org> wrote:

On Thu, Mar 1, 2012 at 4:49 PM, tech@wildintellect.com via RT <
support@osuosl.org> wrote:

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:
> On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org
>wrote:
>
>> The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
>> even though we currently don't use all of it, I think it's a good idea
>> being prepared to do so.
>>
>
> That's going to take a nice chunk of space on our server but
> its manageable. Does your project have any funding to offset the costs
of
> having us do the backups? I can give you an estimate for that if you'd
like.
>

Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever units
make sense. The alternative we've been considering is buying a new server.

So I just re-did all our "rates" recently to figure out costs and it boils
down to the following; $50/client up to 500G per client. So that comes up
to $500/mo or $6,000/yr and covers around 5T of data. This includes staff
time to setup and maintain the system and replace the hardware in three
years. But I'm obviously flexible as this is just an estimate.

A copy of our current bacula director config has been sent to Lance. In
some cases (e.g. Apache) we actually want to keep some logs. We are
still undergoing some refinement to make the whole thing more efficient.

Perfect!

I assume you can take care of the keys, since you have access to all the

guests.

Yup! Are you currently managing the bacula configs with configuration
management on your end? I don't want us to run into that.

> *nods* Currently restores have to be initiated by us but we are working
> towards creating access to work around that. Will this be a problem?
>
> Thanks-
>
Generally no, we hope to never have a need for such a restore and in
such an event we'd likely be filing a ticket with you anyways because
something went wrong with the hardware or VMs. Though access to extract
some files from the backups would be nice long term.

*nods*

> That's going to take a nice chunk of space on our server but its

We're more than happy to buy disks if that's the best solution.

The cost above generally covers most cases and the server itself is
already full of disks.

Thanks-

--
Lance Albertson
Systems Administrator / Architect Open Source Lab
Information Services Oregon State University

16

I think I've got multiple tickets related to this topic open now.

Yes, we do still need to solve our backup needs. The current idea from
OSGeo is a 1U server with 4x2TB or 4x3TB drives for $1500-$2500. That
gives us 4-6 TB RAID 6 usable and should last us 3-5 years.

If you guys can match that pricing and level of service we could use
your service. Otherwise we'd love a quote on hardware we can pay for to
get something in soon.

Thanks,
Alex

On 10/20/2012 03:32 PM, Lance Albertson via RT wrote:

We're you still want to do this? We can always work out a better deal if
the cost is too high.

Thanks
On Mar 1, 2012 5:07 PM, "Lance Albertson" <ramereth@osuosl.org> wrote:

On Thu, Mar 1, 2012 at 4:49 PM, tech@wildintellect.com via RT <
support@osuosl.org> wrote:

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT <support@osuosl.org
wrote:

The combined net disk space on the OSGeo VM's is approx. 2 TByte and,
even though we currently don't use all of it, I think it's a good idea
being prepared to do so.

That's going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the costs

of

having us do the backups? I can give you an estimate for that if you'd

like.

Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever units
make sense. The alternative we've been considering is buying a new server.

So I just re-did all our "rates" recently to figure out costs and it boils
down to the following; $50/client up to 500G per client. So that comes up
to $500/mo or $6,000/yr and covers around 5T of data. This includes staff
time to setup and maintain the system and replace the hardware in three
years. But I'm obviously flexible as this is just an estimate.

A copy of our current bacula director config has been sent to Lance. In
some cases (e.g. Apache) we actually want to keep some logs. We are
still undergoing some refinement to make the whole thing more efficient.

  Perfect!

I assume you can take care of the keys, since you have access to all the

guests.

Yup! Are you currently managing the bacula configs with configuration
management on your end? I don't want us to run into that.

  > *nods* Currently restores have to be initiated by us but we are working

towards creating access to work around that. Will this be a problem?

Thanks-

Generally no, we hope to never have a need for such a restore and in
such an event we'd likely be filing a ticket with you anyways because
something went wrong with the hardware or VMs. Though access to extract
some files from the backups would be nice long term.

*nods*

That's going to take a nice chunk of space on our server but its

We're more than happy to buy disks if that's the best solution.

The cost above generally covers most cases and the server itself is
already full of disks.

Thanks-

--
Lance Albertson
Systems Administrator / Architect Open Source Lab
Information Services Oregon State University

17

Hello Alex,

Given the large storage requirements, we feel that you may be better off purchasing and managing your own server for now. The server specs are largely up to you, since what you'll need is heavily dependent on what you plan to do with it. We would, however, like to see what you have decided on before you order it. Let us know if you have any questions.

Thanks,
Sean

---

Sean Rettig
Student Systems Engineer
OSU Open Source Lab

On Wed Mar 20 10:53:41 2013, tech@wildintellect.com wrote:

I think I've got multiple tickets related to this topic open now.

Yes, we do still need to solve our backup needs. The current idea from
OSGeo is a 1U server with 4x2TB or 4x3TB drives for $1500-$2500. That
gives us 4-6 TB RAID 6 usable and should last us 3-5 years.

If you guys can match that pricing and level of service we could use
your service. Otherwise we'd love a quote on hardware we can pay for
to
get something in soon.

Thanks,
Alex

On 10/20/2012 03:32 PM, Lance Albertson via RT wrote:
> We're you still want to do this? We can always work out a better
deal if
> the cost is too high.
>
> Thanks
> On Mar 1, 2012 5:07 PM, "Lance Albertson" <ramereth@osuosl.org>
wrote:
>
>>
>>
>> On Thu, Mar 1, 2012 at 4:49 PM, tech@wildintellect.com via RT <
>> support@osuosl.org> wrote:
>>
>>> On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:
>>>> On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT
<support@osuosl.org
>>>> wrote:
>>>>
>>>>> The combined net disk space on the OSGeo VM's is approx. 2 TByte
and,
>>>>> even though we currently don't use all of it, I think it's a
good idea
>>>>> being prepared to do so.
>>>>>
>>>>
>>>> That's going to take a nice chunk of space on our server but
>>>> its manageable. Does your project have any funding to offset the
costs
>>> of
>>>> having us do the backups? I can give you an estimate for that if
you'd
>>> like.
>>>>
>>>
>>> Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever
units
>>> make sense. The alternative we've been considering is buying a new
server.
>>
>>
>> So I just re-did all our "rates" recently to figure out costs and
it boils
>> down to the following; $50/client up to 500G per client. So that
comes up
>> to $500/mo or $6,000/yr and covers around 5T of data. This includes
staff
>> time to setup and maintain the system and replace the hardware in
three
>> years. But I'm obviously flexible as this is just an estimate.
>>
>>
>>> A copy of our current bacula director config has been sent to
Lance. In
>>> some cases (e.g. Apache) we actually want to keep some logs. We
are
>>> still undergoing some refinement to make the whole thing more
efficient.
>>>
>>
>> Perfect!
>>
>> I assume you can take care of the keys, since you have access to
all the
>>> guests.
>>
>>
>> Yup! Are you currently managing the bacula configs with
configuration
>> management on your end? I don't want us to run into that.
>>
>>
>>> > *nods* Currently restores have to be initiated by us but we
are working
>>>> towards creating access to work around that. Will this be a
problem?
>>>>
>>>> Thanks-
>>>>
>>> Generally no, we hope to never have a need for such a restore and
in
>>> such an event we'd likely be filing a ticket with you anyways
because
>>> something went wrong with the hardware or VMs. Though access to
extract
>>> some files from the backups would be nice long term.
>>
>>
>> *nods*
>>
>>> That's going to take a nice chunk of space on our server but its
>>>
>>> We're more than happy to buy disks if that's the best solution.
>>>
>>
>> The cost above generally covers most cases and the server itself is
>> already full of disks.
>>
>> Thanks-
>>
>> --
>> Lance Albertson
>> Systems Administrator / Architect Open
Source Lab
>> Information Services Oregon State
University
>>
>>
>

18

Do you still have a strong preference for Dell? I've been having
unpleasant ordering experiences with them, along with pricing
challenges. Personally I've had a better experience with Silicon
Mechanics which is pretty local to you.

We're looking at a 1U 4 disk system (2, 3 or 4 TB drives). 1-2 lower end
sockets.

I'll put together some quotes.

Thanks,
Alex

On 08/02/2013 10:23 AM, Sean Rettig via RT wrote:

Hello Alex,

Given the large storage requirements, we feel that you may be better off purchasing and managing your own server for now. The server specs are largely up to you, since what you'll need is heavily dependent on what you plan to do with it. We would, however, like to see what you have decided on before you order it. Let us know if you have any questions.

Thanks,
Sean

---

Sean Rettig
Student Systems Engineer
OSU Open Source Lab

On Wed Mar 20 10:53:41 2013, tech@wildintellect.com wrote:

I think I've got multiple tickets related to this topic open now.

Yes, we do still need to solve our backup needs. The current idea from
OSGeo is a 1U server with 4x2TB or 4x3TB drives for $1500-$2500. That
gives us 4-6 TB RAID 6 usable and should last us 3-5 years.

If you guys can match that pricing and level of service we could use
your service. Otherwise we'd love a quote on hardware we can pay for
to
get something in soon.

Thanks,
Alex

On 10/20/2012 03:32 PM, Lance Albertson via RT wrote:

We're you still want to do this? We can always work out a better

deal if

the cost is too high.

Thanks
On Mar 1, 2012 5:07 PM, "Lance Albertson" <ramereth@osuosl.org>

wrote:

On Thu, Mar 1, 2012 at 4:49 PM, tech@wildintellect.com via RT <
support@osuosl.org> wrote:

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT

<support@osuosl.org

wrote:

The combined net disk space on the OSGeo VM's is approx. 2 TByte

and,

even though we currently don't use all of it, I think it's a

good idea

being prepared to do so.

That's going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the

costs

of

having us do the backups? I can give you an estimate for that if

you'd

like.

Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever

units

make sense. The alternative we've been considering is buying a new

server.

So I just re-did all our "rates" recently to figure out costs and

it boils

down to the following; $50/client up to 500G per client. So that

comes up

to $500/mo or $6,000/yr and covers around 5T of data. This includes

staff

time to setup and maintain the system and replace the hardware in

three

years. But I'm obviously flexible as this is just an estimate.

A copy of our current bacula director config has been sent to

Lance. In

some cases (e.g. Apache) we actually want to keep some logs. We

are

still undergoing some refinement to make the whole thing more

efficient.

  Perfect!

I assume you can take care of the keys, since you have access to

all the

guests.

Yup! Are you currently managing the bacula configs with

configuration

management on your end? I don't want us to run into that.

  > *nods* Currently restores have to be initiated by us but we

are working

towards creating access to work around that. Will this be a

problem?

Thanks-

Generally no, we hope to never have a need for such a restore and

in

such an event we'd likely be filing a ticket with you anyways

because

something went wrong with the hardware or VMs. Though access to

extract

some files from the backups would be nice long term.

*nods*

That's going to take a nice chunk of space on our server but its

We're more than happy to buy disks if that's the best solution.

The cost above generally covers most cases and the server itself is
already full of disks.

Thanks-

--
Lance Albertson
Systems Administrator / Architect Open

Source Lab

Information Services Oregon State

University

19

Do you still have a strong preference for Dell? I've been having
unpleasant ordering experiences with them, along with pricing
challenges. Personally I've had a better experience with Silicon
Mechanics which is pretty local to you.

We're looking at a 1U 4 disk system (2, 3 or 4 TB drives). 1-2 lower end
sockets.

I'll put together some quotes.

Thanks,
Alex

On 08/02/2013 10:23 AM, Sean Rettig via RT wrote:

Hello Alex,

Given the large storage requirements, we feel that you may be better off purchasing and managing your own server for now. The server specs are largely up to you, since what you'll need is heavily dependent on what you plan to do with it. We would, however, like to see what you have decided on before you order it. Let us know if you have any questions.

Thanks,
Sean

---

Sean Rettig
Student Systems Engineer
OSU Open Source Lab

On Wed Mar 20 10:53:41 2013, tech@wildintellect.com wrote:

I think I've got multiple tickets related to this topic open now.

Yes, we do still need to solve our backup needs. The current idea from
OSGeo is a 1U server with 4x2TB or 4x3TB drives for $1500-$2500. That
gives us 4-6 TB RAID 6 usable and should last us 3-5 years.

If you guys can match that pricing and level of service we could use
your service. Otherwise we'd love a quote on hardware we can pay for
to
get something in soon.

Thanks,
Alex

On 10/20/2012 03:32 PM, Lance Albertson via RT wrote:

We're you still want to do this? We can always work out a better

deal if

the cost is too high.

Thanks
On Mar 1, 2012 5:07 PM, "Lance Albertson" <ramereth@osuosl.org>

wrote:

On Thu, Mar 1, 2012 at 4:49 PM, tech@wildintellect.com via RT <
support@osuosl.org> wrote:

On 03/01/2012 04:35 PM, Lance Albertson via RT wrote:

On Thu, Mar 1, 2012 at 6:47 AM, Martin Spott via RT

<support@osuosl.org

wrote:

The combined net disk space on the OSGeo VM's is approx. 2 TByte

and,

even though we currently don't use all of it, I think it's a

good idea

being prepared to do so.

That's going to take a nice chunk of space on our server but
its manageable. Does your project have any funding to offset the

costs

of

having us do the backups? I can give you an estimate for that if

you'd

like.

Yes, we'd like an estimate per GB, or 100 GB or 1 TB. Whichever

units

make sense. The alternative we've been considering is buying a new

server.

So I just re-did all our "rates" recently to figure out costs and

it boils

down to the following; $50/client up to 500G per client. So that

comes up

to $500/mo or $6,000/yr and covers around 5T of data. This includes

staff

time to setup and maintain the system and replace the hardware in

three

years. But I'm obviously flexible as this is just an estimate.

A copy of our current bacula director config has been sent to

Lance. In

some cases (e.g. Apache) we actually want to keep some logs. We

are

still undergoing some refinement to make the whole thing more

efficient.

  Perfect!

I assume you can take care of the keys, since you have access to

all the

guests.

Yup! Are you currently managing the bacula configs with

configuration

management on your end? I don't want us to run into that.

  > *nods* Currently restores have to be initiated by us but we

are working

towards creating access to work around that. Will this be a

problem?

Thanks-

Generally no, we hope to never have a need for such a restore and

in

such an event we'd likely be filing a ticket with you anyways

because

something went wrong with the hardware or VMs. Though access to

extract

some files from the backups would be nice long term.

*nods*

That's going to take a nice chunk of space on our server but its

We're more than happy to buy disks if that's the best solution.

The cost above generally covers most cases and the server itself is
already full of disks.

Thanks-

--
Lance Albertson
Systems Administrator / Architect Open

Source Lab

Information Services Oregon State

University

20

On Fri, Aug 2, 2013 at 10:45 AM, tech@wildintellect.com via RT <
support@osuosl.org> wrote:

Do you still have a strong preference for Dell? I've been having
unpleasant ordering experiences with them, along with pricing
challenges. Personally I've had a better experience with Silicon
Mechanics which is pretty local to you.

We're looking at a 1U 4 disk system (2, 3 or 4 TB drives). 1-2 lower end
sockets.

I'll put together some quotes.

Ya I totally recommend Silicon Mechanics. We'll take a look at the quote
once you send it.

--
Lance Albertson
Director
Oregon State University | Open Source Lab