interrupted external restore and data availability
Hi,
we have a serious issue with a 420GB instance we need to restore. Actually, we
only need a table from one of the databases it contains, but we cannot afford
to restore it on the production environment (it takes over 40 hrs). We usually
restore the instance via an external restore on a dedicated server and access
the data we need.
My question is, since we only need a small part of the the instance, is it
possible to stop the restore (onbar) as soon as the bar_act.log says the
dbspace we need has been restored and forget about the rest of dbspaces (as
long as the critical dbspaces have been restored too, obviously)? When we then
perform an onmode -m to put it in On-Line mode, are the dbspaces whose restore
have been completed accessible or is everything lost and we have to let the
onbar restore process finish?
For the record, we restore with onbar -r -p (only physical restore). And it's
an IDS 9.40.FC6 on solaris 9, database storage on raw devices.
BTW, we also tried table level restore (archecker) from an IDS 10.00, but it
didn't quite finish: we reached a TIMEOUT (5hrs,30min) and we would like to
repeat this with a higher timeout, but the second question is, is this
possible? I mean, can an IDS 10.00 via archecker access the data stored with
onbar by an IDS 9.40 (both of them solaris 9), or is there a binary
incompatibility (like when you try, e.g. to restore from an ontape copy
performed with an IDS 9.40 on an IDS 10.00 platform)?
Thanks a lot in advance!
Bye
I just had to do this on a similarily sized instance on Friday. You want to
use archecker to recover only the table. Shoot me An email address to
dan.mueller@trnswrks.com and I will send you a great cheat sheet.
Dan
-----Original Message-----
From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of GERARDO
PADIERNA
Sent: Monday, February 28, 2011 3:30 PM
To: ids@iiug.org
Subject: interrupted external restore and data availability [22945]
Hi,
we have a serious issue with a 420GB instance we need to restore. Actually, we
only need a table from one of the databases it contains, but we cannot afford
to restore it on the production environment (it takes over 40 hrs). We usually
restore the instance via an external restore on a dedicated server and access
the data we need.
My question is, since we only need a small part of the the instance, is it
possible to stop the restore (onbar) as soon as the bar_act.log says the
dbspace we need has been restored and forget about the rest of dbspaces (as
long as the critical dbspaces have been restored too, obviously)? When we then
perform an onmode -m to put it in On-Line mode, are the dbspaces whose restore
have been completed accessible or is everything lost and we have to let the
onbar restore process finish?
For the record, we restore with onbar -r -p (only physical restore). And it's
an IDS 9.40.FC6 on solaris 9, database storage on raw devices.
BTW, we also tried table level restore (archecker) from an IDS 10.00, but it
didn't quite finish: we reached a TIMEOUT (5hrs,30min) and we would like to
repeat this with a higher timeout, but the second question is, is this
possible? I mean, can an IDS 10.00 via archecker access the data stored with
onbar by an IDS 9.40 (both of them solaris 9), or is there a binary
incompatibility (like when you try, e.g. to restore from an ontape copy
performed with an IDS 9.40 on an IDS 10.00 platform)?
Thanks a lot in advance!
Bye
*******************************************************************************
Forum Note: Use "Reply" to post a response in the discussion forum.
Actually here it is and once again, my thanx to joe plugge for helping me out
on the same issue.
Dan
Create a command file that contains the schema of the table from and table to
(don't forget to put what dbspace(s) it was built in) See note below for
replicated tables and the workaround so that you do not run into a bug that
causes duplicates and other odd behavior due to replicated tables.
Example file (gmc_lows.cmd)
--------------------------------------------------------------------------------
-------------------
database class;
create table gmc_lowes
(
count integer,
card_num char(19) not null ,
order_num char(15) not null ,
pass_code char(26) not null ,
amount char(15) not null ,
expdate char(8) not null ,
stdate char(8) not null ,
sttime char(6) not null ,
status smallint,
call_time integer
) in dbs1 ;
create table gmc_lowes_restore
(
count integer,
card_num char(19) not null ,
order_num char(15) not null ,
pass_code char(26) not null ,
amount char(15) not null ,
expdate char(8) not null ,
stdate char(8) not null ,
sttime char(6) not null ,
status smallint,
call_time integer
) in dbs1 ;
insert into gmc_lowes_restore select * from gmc_lowes;
restore to '2008-04-03 14:00:00';
--------------------------------------------------------------------------------
-------------------
Here is the command to run:
archecker -bXvs -f /wic/scratch/gmc_lows.cmd
This is what you will see on the screen:
informix @ ibm91l:/wic/scratch: archecker -bXvs -f /wic/scratch/gmc_lows.cmd
IBM Informix Dynamic Server Version 10.00.FC6
Program Name: archecker
Version: 8.0
Released: 2006-12-19 00:20:38
CSDK: IBM Informix CSDK Version 2.90
ESQL: IBM Informix-ESQL Version 2.90.FN458
Compiled: 12/19/06 00:22 on AIX 2 5
AC_STORAGE /tmp
AC_MSGPATH /tmp/ac_msg.log
AC_VERBOSE on
AC_TAPEBLOCK 60 KB
AC_IXBAR /wic/informix/etc/ixbar.91
Dropping old log control tables
Extracting table class:gmc_lowes into class:gmc_lowes_restore
........................................
.....................................
Scan PASSED
Control page checks PASSED
Table checks PASSED
Table extraction commands 1
Tables found on archive 1
LOADED: class:gmc_lowes_restore produced 30 rows.
Creating log control tables
Staging Log 1888
** NOTE *** For Replicated TABLES :
###########################################################################
The cmd file needs to look like this- (note that cdrserver and cdrtime are
both int's):
--------------------------------------------------------------------------------
-------------------
database stores_demo;
create table rocket
(cdrserver int,
cdrtime int,
rocket_id int,
rocket_name char(20),
rocket_cost money(10,2),
launch_date datetime year to second,
primary key (rocket_id,rocket_name)
) in dbs1;
create table rocket2
(
rocket_id int,
rocket_name char(20),
rocket_cost money(10,2),
launch_date datetime year to second,
primary key (rocket_id,rocket_name)
) with crcols;
insert into rocket2 select rocket_id, rocket_name, rocket_cost,launch_date
from rocket; restore to current with no log;--------------------------------------------------------------------------------
-------------------
-----Original Message-----
From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of GERARDO
PADIERNA
Sent: Monday, February 28, 2011 3:30 PM
To: ids@iiug.org
Subject: interrupted external restore and data availability [22945]
Hi,
we have a serious issue with a 420GB instance we need to restore. Actually, we
only need a table from one of the databases it contains, but we cannot afford
to restore it on the production environment (it takes over 40 hrs). We usually
restore the instance via an external restore on a dedicated server and access
the data we need.
My question is, since we only need a small part of the the instance, is it
possible to stop the restore (onbar) as soon as the bar_act.log says the
dbspace we need has been restored and forget about the rest of dbspaces (as
long as the critical dbspaces have been restored too, obviously)? When we then
perform an onmode -m to put it in On-Line mode, are the dbspaces whose restore
have been completed accessible or is everything lost and we have to let the
onbar restore process finish?
For the record, we restore with onbar -r -p (only physical restore). And it's
an IDS 9.40.FC6 on solaris 9, database storage on raw devices.
BTW, we also tried table level restore (archecker) from an IDS 10.00, but it
didn't quite finish: we reached a TIMEOUT (5hrs,30min) and we would like to
repeat this with a higher timeout, but the second question is, is this
possible? I mean, can an IDS 10.00 via archecker access the data stored with
onbar by an IDS 9.40 (both of them solaris 9), or is there a binary
incompatibility (like when you try, e.g. to restore from an ontape copy
performed with an IDS 9.40 on an IDS 10.00 platform)?
Thanks a lot in advance!
Bye
*******************************************************************************
Forum Note: Use "Reply" to post a response in the discussion forum.
Hi Dan,
thanks a lot for your cheat sheet, in fact we already tried that, and it
worked, or at least, it seems to work perfectly: we have a schema file that
starts perfectly well, the tape mounts, archecker starts to fetch the data...
but they never come, at least not for for about 5h33min (20000 seconds). We
got a timeout and I customer sticked to the old style external restore.
My concern with respect to archecker is:
- can an IDS 10 (or 11, or 11.50, whatever) access a binary onbar copy,
performed by an IDS 9.40 instance? According to the IBM documentation it
should work (it's claimed to be a valid tool for migrating between different
platforms)
- is there a way to determine beforehand how long it will take aprox.?
The thing is, our customer is already quite annoyed by the fact that it's
taking so long (we have an external restore just right now running, since
Saturday morning, and the estimations are that it'll take yet another 7
days!!! The fundamental thing here is, they don't have free storage space and
we're working with NFS; the chunks where we're restoring are on an NFS file
system, and that is painfully slow).
If it wasn't so sad I just would LOL!!
Thanks.
Gerardo
See details below.
John F. Miller III
STSM, Embedability Architect
miller3@us.ibm.com
503-578-5645
IBM Informix Dynamic Server (IDS)
ids-bounces@iiug.org wrote on 02/28/2011 02:43:11 PM:
> [image removed]
>
> Re: RE: interrupted external restore and data avai [22948]
>
> GERARDO PADIERNA
>
> to:
>
> ids
>
> 02/28/2011 02:44 PM
>
> Sent by:
>
> ids-bounces@iiug.org
>
> Please respond to ids
>
> Hi Dan,
> thanks a lot for your cheat sheet, in fact we already tried that, and it
> worked, or at least, it seems to work perfectly: we have a schema file
that
> starts perfectly well, the tape mounts, archecker starts to fetch the
data...
> but they never come, at least not for for about 5h33min (20000 seconds).
We
> got a timeout and I customer sticked to the old style external restore.
> My concern with respect to archecker is:
>
> - can an IDS 10 (or 11, or 11.50, whatever) access a binary onbar copy,
> performed by an IDS 9.40 instance? According to the IBM documentation it
> should work (it's claimed to be a valid tool for migrating between
different
> platforms)
The archecker version and the server version which took the archive must be
the same version. What is not version dependant is where the data goes.
The
target tables maybe different informix versions as it does nothing more
than
insert the data.
> - is there a way to determine beforehand how long it will take aprox.?
If you are talking about archecker, it prints out a dot to the command line
(if -v option is used) for every 250MB of data processed.
If you are using onbar then be SURE TO SPECIFIC the dbspaces in the source
tables schema. If you do specific the dbspaces then archecker will only
scan
those dbspaces reducing your restore time. If you do not specific
the dbspaces then archecker will look at ALL dbspaces.
> The thing is, our customer is already quite annoyed by the fact that it's
> taking so long (we have an external restore just right now running, since
> Saturday morning, and the estimations are that it'll take yet another 7
> days!!! The fundamental thing here is, they don't have free storage space
and
> we're working with NFS; the chunks where we're restoring are on an NFS
file
> system, and that is painfully slow).
>
> If it wasn't so sad I just would LOL!!
> Thanks.
>
> Gerardo
>
>
>
*******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>