Andrey Gorchivkin wrote:
Thanks a lot for your response Rafiu and keep up the good work :).

Although the product is progressing very well and I was able to get back the 
iSCSI functionality after a clean install, I'm still missing some components 
from it.

1. Full VMware-Tools support - it would be nice if you or someone else can post 
at least a how to on manually compiling and integrating this package inside OF.

Have you looked in the forums for details about this? http://www.openfiler.com/community/forums. IIRC it's been documented somewhere before.


 I didn’t succeed on that approach unfortunately. The best would be if 
VMware-Tools is integrated via Conary.

Can't do that due to licensing issues. We *may* add it to subscription-only repository though.

2. It will be very handy if OpenFile can create iSCSI volumes on file storage 
basis (not only on disks). This would make it possible to fully utilize the 
iSCSI based snapshots and by the way this is the way how big storage companies 
like NetApp are solving the same situation. Disk based iSCSI volumes would 
still remain very useful but it would be very nice to have the other option 
available too.

Agreed. We'll implement in OF 2.4.


3. As we are talking about iSCSI snapshots is there a comfortable and safe way 
on restoring them without hearting the data integrity ? I already use OF boxes 
as a shared storage for my Exchange, SQL and NetBackup clusters but I'm afraid 
that I won't be able to quickly restore crucial iSCSI volumes when needed. 
Please advise us on that topic as possible.


Setting up access for an iSCSI snapshot is as easy as enabling it via ietadm command line tool.


3. With my clusters in the infrastructure sometimes it's difficult to track the iSCSI volumes throughout cluster nodes, because of their amount

How many nodes/iSCSI luns? (just out of curiousity)


and especially when more than one OF box is being used. In such cases the 
non-unique iSCSI target name per OF box can become hell. Would it be possible 
to place also an option for changing the target name from the web interface ?
Hmm, I've been giving this some thought recently for another product I'm working on. And the conclusion I've come to is that we have uuid based names - for uniqueness, and an option to add an alias (that would be the name of the target you enter in the webform). We can't obviously mix and match as it would be a nightmare to manage.


4. This one is more about the documentation and guides section. As I'm doing my 
clusters I had very hard time on inventing a good naming standard for the 
software raid based iSCSI volumes. To be able identifying volumes easily I 
implemented a special naming concept in the form of (RxPVxxiVxx) where the R, 
PV, and iV are indicating the different layers from RAID volumes to iSCSI like 
this:

Disk Layer      -       no naming here  -       250GB HDD x 4
RAID Layer      -       no naming here  -       RAID-5 + spare
PV Layer        -       R5PV01          -       Rx indicates currently used 
RAID level while PVxx indicates the physical volume
VG Layer        -       R5PV01iV01              -       To avoid any duplicates 
at this time I included the PV identification here too

So all in all OpenFiler machine with 1 x RAID-5 and 2 x RAID-10 + 3 iSCSI 
volumes on each RAID would look like this:

RAID-5 Config

PV Layer        -       R5PV01
VG Layer        -       R5PV01iV01
VG Layer        -       R5PV01iV02
VG Layer        -       R5PV01iV03

RAID-10 Config #1

PV Layer        -       R10PV01
VG Layer        -       R10PV01iV01
VG Layer        -       R10PV01iV02
VG Layer        -       R10PV01iV03

RAID-10 Config #1

PV Layer        -       R10PV02
VG Layer        -       R10PV02iV01
VG Layer        -       R10PV02iV02
VG Layer        -       R10PV02iV03

With this naming standard it was pretty easy to identify and track the exact 
iSCSI volume for every single node of my clusters. I just thought this might be 
included as hint in the Web Interface or at least as part of the documentation 
and guides like an example.

We could stick it somewhere in the wiki.

5. Some other points regarding iSCSI (and other) volumes in OF would be the grouping option. Again with my clusters it was a bit frustrating to go back all the time and check the whole list of volumes for the correct iSCSI record.
If we have UUID names, this shouldn't be an issue.

It will be very useful and nice if we can create customizable groups for the different 
set of volumes. In my case I wanted to group the volumes on application basis (i.e. SQL 
cluster is assigned with 3 volumes for Quorum, MSDTC and Database). Currently I'm using 
the provided "Comments" field for that purpose but the chance for accidentally 
messing things up is pretty big that way.

Please elaborate further here if you can....



R.

Regards,

Andrey

-----Original Message-----
From: Rafiu M. Fakunle [mailto:[EMAIL PROTECTED] Sent: Tuesday, May 29, 2007 11:38 PM
To: Andrey Gorchivkin
Cc: [email protected]
Subject: Re: [OF-users] OpenFiler missing components ???

Hi,
----- "Andrey Gorchivkin" <[EMAIL PROTECTED]> wrote:
Hi All,

Yesterday I tried to update one of my OF boxes and I had to realize
that
many packages were removed after "conary updateall". I even tried the
update
on a pure installation, but the result was the same. At this time
iscsi-target, RAA and several other packages cannot be used at all
which
makes OF pretty useless.


So obviously my questions are:

1.       Were those packages been removed on purpose ?

iscsi-trgt(-kernel): yes. RAA: inadvertently.
2.       When can we expect the availability of the same packages
again ?

iscsi_trgt-kernel is no longer required as we've integrated the module (along with drbd) into the shipped kernel. Makes for easier update management. RAA changed in conary and I missed the announcement. New group builds have the updated references.
3.       Global announcement regarding such changes would be very
useful to
all OF users even if the software is free. Can you implement such
announcement for the future ? (perhaps on this mailing list or main
web
site)

Yes. We'll do that.

--

R.


_______________________________________________
Openfiler-users mailing list
[email protected]
https://lists.openfiler.com/mailman/listinfo/openfiler-users

Reply via email to