[Info-Ingres] Ingres and Dell-Equalogic SAN

[Database Administrator]

Hi All,



I'm trying to make Ingres do backup and recoveries on a Dell Equalogic
SAN.



It looks like a simple matter of creating and restoring snapshots, by
monkeying with the right points of the cktmpl.def.



It seems simple enough to do the backup, but the restore is a tad more
entertaining. In particular the Dell Manual states:

Before restoring a volume from a snapshot, be sure to disconnect any
iSCSI initiators from the volume. Otherwise, the restored volume may be
corrupted when initiators reconnect to it.



My plan on recovery was to get the tape based template commands to do
the job. So I'd have to get the WRTD: to do the deed probably via a
script to ensure we disconnect the iSCSI, and then do the restore from
the snapshot.



Does anyone have an idea on how to do this from a script? Or should I
just get the PRTD to echo a message prompting me to do this and not
continuing until the RETURN key is hit in the WRTD?



Martin Bowes

[Database Administrator]

On Nov 3, 10:08*am, "Martin Bowes" wrote:
> Hi All,
>
> I'm trying to make Ingres do backup and recoveries on a Dell Equalogic
> SAN.
>
> It looks like a simple matter of creating and restoring snapshots, by
> monkeying with the right points of the cktmpl.def.
>
> It seems simple enough to do the backup, but the restore is a tad more
> entertaining. In particular the Dell Manual states:
>
> Before restoring a volume from a snapshot, be sure to disconnect any
> iSCSI initiators from the volume. Otherwise, the restored volume may be
> corrupted when initiators reconnect to it.
>
> My plan on recovery was to get the tape based template commands to do
> the job. So I'd have to get the WRTD: to do the deed probably via a
> script to ensure we disconnect the iSCSI, and then do the restore from
> the snapshot.
>
> Does anyone have an idea on how to do this from a script? Or should I
> just get the PRTD to echo a message prompting me to do this and not
> continuing until the RETURN key is hit in the WRTD?
>
> Martin Bowes

Marty,

I believe you will find command(s) may be unique to the O/Ses
connected to the array, rather than the array itself. That being the
case, who is talking to it, Windoooz, Linux, or ????

Scott

[Database Administrator]

http://www.equallogic.com/uploadedFi...n-Examples.pdf

may have some entertainment value as well

[Database Administrator]

Hi Scott,

I decided to hand the responsibility of disconnecting the iSCSI initiators back to the systems dudes...they looked a lot happier when I said that...

So I'll just a prompt in the PRTD and continue in the WRTD assuming its all been done. Much like the existing PRTD: echo 'please mount tape..' and WRTD: read foo;...

Marty

-----Original Message-----
From: info-ingres-bounces@kettleriverconsulting.com [mailto:info-ingres-bounces@kettleriverconsulting.com] On Behalf Of OldSchool
Sent: 03 November 2008 15:58
To: info-ingres@kettleriverconsulting.com
Subject: Re: [Info-Ingres] Ingres and Dell-Equalogic SAN

On Nov 3, 10:08*am, "Martin Bowes" wrote:
> Hi All,
>
> I'm trying to make Ingres do backup and recoveries on a Dell Equalogic
> SAN.
>
> It looks like a simple matter of creating and restoring snapshots, by
> monkeying with the right points of the cktmpl.def.
>
> It seems simple enough to do the backup, but the restore is a tad more
> entertaining. In particular the Dell Manual states:
>
> Before restoring a volume from a snapshot, be sure to disconnect any
> iSCSI initiators from the volume. Otherwise, the restored volume may be
> corrupted when initiators reconnect to it.
>
> My plan on recovery was to get the tape based template commands to do
> the job. So I'd have to get the WRTD: to do the deed probably via a
> script to ensure we disconnect the iSCSI, and then do the restore from
> the snapshot.
>
> Does anyone have an idea on how to do this from a script? Or should I
> just get the PRTD to echo a message prompting me to do this and not
> continuing until the RETURN key is hit in the WRTD?
>
> Martin Bowes

Marty,

I believe you will find command(s) may be unique to the O/Ses
connected to the array, rather than the array itself. That being the
case, who is talking to it, Windoooz, Linux, or ????

Scott
_______________________________________________
Info-Ingres mailing list
Info-Ingres@kettleriverconsulting.com
http://www.kettleriverconsulting.com...fo/info-ingres

[Database Administrator]

Hi Scott,

Thanks for that. I'll have a read shortly.

Marty

-----Original Message-----
From: info-ingres-bounces@kettleriverconsulting.com
[mailto:info-ingres-bounces@kettleriverconsulting.com] On Behalf Of
OldSchool
Sent: 03 November 2008 16:01
To: info-ingres@kettleriverconsulting.com
Subject: Re: [Info-Ingres] Ingres and Dell-Equalogic SAN

http://www.equallogic.com/uploadedFi...orts/TR1038-Da
ta-Protection-Examples.pdf

may have some entertainment value as well

_______________________________________________
Info-Ingres mailing list
Info-Ingres@kettleriverconsulting.com
http://www.kettleriverconsulting.com...fo/info-ingres

[Database Administrator]

On Nov 3, 10:08*am, "Martin Bowes" wrote:
> Hi All,
>
> I'm trying to make Ingres do backup and recoveries on a Dell Equalogic
> SAN.
>
> It looks like a simple matter of creating and restoring snapshots, by
> monkeying with the right points of the cktmpl.def.
>
> It seems simple enough to do the backup, but the restore is a tad more
> entertaining. In particular the Dell Manual states:
>
> Before restoring a volume from a snapshot, be sure to disconnect any
> iSCSI initiators from the volume. Otherwise, the restored volume may be
> corrupted when initiators reconnect to it.
>
> My plan on recovery was to get the tape based template commands to do
> the job. So I'd have to get the WRTD: to do the deed probably via a
> script to ensure we disconnect the iSCSI, and then do the restore from
> the snapshot.
>
> Does anyone have an idea on how to do this from a script? Or should I
> just get the PRTD to echo a message prompting me to do this and not
> continuing until the RETURN key is hit in the WRTD?
>
> Martin Bowes

So...here's a silly question: In such an environment, how does one go
about insuring that the DB is consistent (buffers flushed, everything
committed and so forth). I've been shutting down the installation(s)
for the time required to perform the snapshots, but is there a better
way.

[Database Administrator]

On Nov 4, 2008, at 9:19 AM, OldSchool wrote:

> On Nov 3, 10:08 am, "Martin Bowes" wrote:
>> Hi All,
>>
>> I'm trying to make Ingres do backup and recoveries on a Dell
>> Equalogic
>> SAN.
>>
>> It looks like a simple matter of creating and restoring snapshots, by
>> monkeying with the right points of the cktmpl.def.
>>
>> [snip]
> So...here's a silly question: In such an environment, how does one go
> about insuring that the DB is consistent (buffers flushed, everything
> committed and so forth). I've been shutting down the installation(s)
> for the time required to perform the snapshots, but is there a better
> way.
>

You basically have two choices. You can tie the SAN snapshot into an
off-line checkpoint, which is certainly the easiest way. The off-line
checkpoint arranges to do a consistency point ahead of time, which
flushes buffers, and prohibits any changes during the checkpoint.
This is easiest partly because restoration just involves restoring
the snapshot, very little trickiness involved.

The other way is to tie the SAN snapshot to an online checkpoint, and
preserve the dump area too. The tricky part here is that to restore
the snapshot, you have to do a real rollforward. Since rollforwarddb
is going to insist on deleting the contents of the data area first,
integrating the rollforward with the snapshot restore is a bit
harder -- and you have to have someplace to put the dump area stuff,
and tie it to the proper snapshots, and get it put back into the
right places at the right time. It's definitely more work to get right.

Karl

[Database Administrator]

For the backup, you are guarunteed that the database has been made consistent on disk at the point given in the template file by the BSTD. At this point all user activity is stalled, cacahe buffers written to disks.

So all I've got in my cktmpl.def is...BSTD: /user/ingres/scripts/SAN/SAN_Snapshot.sh %F %C

The script essentially does:
$SSH_COMMAND snapcol create $volume_collection_name $ckpdb_num \
description "Ingres backup $ ckpdb_num"
exit $?; # return status of snapcol create command

So my snapshot is named after the Ingres ckpdb number.

Marty

-----Original Message-----
From: info-ingres-bounces@kettleriverconsulting.com [mailto:info-ingres-bounces@kettleriverconsulting.com] On Behalf Of OldSchool
Sent: 04 November 2008 14:19
To: info-ingres@kettleriverconsulting.com
Subject: Re: [Info-Ingres] Ingres and Dell-Equalogic SAN

On Nov 3, 10:08*am, "Martin Bowes" wrote:
> Hi All,
>
> I'm trying to make Ingres do backup and recoveries on a Dell Equalogic
> SAN.
>
> It looks like a simple matter of creating and restoring snapshots, by
> monkeying with the right points of the cktmpl.def.
>
> It seems simple enough to do the backup, but the restore is a tad more
> entertaining. In particular the Dell Manual states:
>
> Before restoring a volume from a snapshot, be sure to disconnect any
> iSCSI initiators from the volume. Otherwise, the restored volume may be
> corrupted when initiators reconnect to it.
>
> My plan on recovery was to get the tape based template commands to do
> the job. So I'd have to get the WRTD: to do the deed probably via a
> script to ensure we disconnect the iSCSI, and then do the restore from
> the snapshot.
>
> Does anyone have an idea on how to do this from a script? Or should I
> just get the PRTD to echo a message prompting me to do this and not
> continuing until the RETURN key is hit in the WRTD?
>
> Martin Bowes

So...here's a silly question: In such an environment, how does one go
about insuring that the DB is consistent (buffers flushed, everything
committed and so forth). I've been shutting down the installation(s)
for the time required to perform the snapshots, but is there a better
way.
_______________________________________________
Info-Ingres mailing list
Info-Ingres@kettleriverconsulting.com
http://www.kettleriverconsulting.com...fo/info-ingres

[Database Administrator]

Hi Karl et al,

The deletion phase of the recovery can be controlled by the WDTD flag.
So rather than do a delete of the data directory, I was simply planning
to do:
WDTD: /bin/echo At WDTD: Do nothing

The snapshots can be recovered at the WRTD, just like a normal tape
restore.

So I think that as long as the dumps and journals are present in the
correct locations then everything else should follow.

The real problem is tieing the checkpoint number to a specific snapshot
name and allowing for a multi-volume set. So I'm having to parse the
output of various SAN commands in a recovery script.

At the moment the problem is getting the SSH key exchange to work. It
just aint. Thus all my SSH commands fail...which is a fairly limiting
feature!

Andre has a solution where he initiates a job on the SAN, but given in
this case the SAN aint running Linux or anything else I can program
with, I'm still stuck.

Marty


-----Original Message-----
From: info-ingres-bounces@kettleriverconsulting.com
[mailto:info-ingres-bounces@kettleriverconsulting.com] On Behalf Of Karl
& Betty Schendel
Sent: 04 November 2008 14:34
To: Ingres and related product discussion forum
Subject: Re: [Info-Ingres] Ingres and Dell-Equalogic SAN


On Nov 4, 2008, at 9:19 AM, OldSchool wrote:

> On Nov 3, 10:08 am, "Martin Bowes" wrote:
>> Hi All,
>>
>> I'm trying to make Ingres do backup and recoveries on a Dell
>> Equalogic
>> SAN.
>>
>> It looks like a simple matter of creating and restoring snapshots, by
>> monkeying with the right points of the cktmpl.def.
>>
>> [snip]
> So...here's a silly question: In such an environment, how does one go
> about insuring that the DB is consistent (buffers flushed, everything
> committed and so forth). I've been shutting down the installation(s)
> for the time required to perform the snapshots, but is there a better
> way.
>

You basically have two choices. You can tie the SAN snapshot into an
off-line checkpoint, which is certainly the easiest way. The off-line
checkpoint arranges to do a consistency point ahead of time, which
flushes buffers, and prohibits any changes during the checkpoint.
This is easiest partly because restoration just involves restoring
the snapshot, very little trickiness involved.

The other way is to tie the SAN snapshot to an online checkpoint, and
preserve the dump area too. The tricky part here is that to restore
the snapshot, you have to do a real rollforward. Since rollforwarddb
is going to insist on deleting the contents of the data area first,
integrating the rollforward with the snapshot restore is a bit
harder -- and you have to have someplace to put the dump area stuff,
and tie it to the proper snapshots, and get it put back into the
right places at the right time. It's definitely more work to get right.

Karl

_______________________________________________
Info-Ingres mailing list
Info-Ingres@kettleriverconsulting.com
http://www.kettleriverconsulting.com...fo/info-ingres

[Database Administrator]

On Nov 6, 2008, at 5:37 AM, Martin Bowes wrote:

> Hi Karl et al,
>
> The deletion phase of the recovery can be controlled by the WDTD flag.
> So rather than do a delete of the data directory, I was simply
> planning
> to do:
> WDTD: /bin/echo At WDTD: Do nothing

That's useful. I assume you've tested this... :-)
Once upon a time, the WDTD only affected added locations for a database,
it would still unconditionally delete the files from the root location!
Which was pretty stupid, of course. I guess that got fixed. That last
time I played with SAN checkpointing in anger was probably Ingres 2.5-
ish.

> At the moment the problem is getting the SSH key exchange to work. It
> just aint. Thus all my SSH commands fail...which is a fairly limiting
> feature!

IMHO, the whole SSH thing is the modern day version of the modem.
It's an invention of the devil, meant to produce frustration... :-)

Karl