Loading master database from backup

[John Flynn]

Hi.

The Troubleshooting Guide in ASE11 and 12 used to have a section called
Encyclopedia of Tasks, and it included a section on loading the master
database from backup. It described a simple three-step process: restart the
database in single-user mode (with -m), issue a "load database master"
command, then restart the database in multi-user mode.

The Troubleshooting Guide in ASE15 seems to lack this section. Instead,
chapter 13 of the System Administration Guide describes a very complicated
procedure for recovering master, including rebuilding the master device and
all the databases residing on it. It seems to make no mention of the simple
three-step process. (When I look back at the System Administration Guide for
ASE11 and 12, I see the same thing, though I never noticed that before.)
Furthermore, the -m option is called master-recover mode (even back in the
SAG for ASE11 and 12). However, the Utility Commands Reference in ASE15
still calls it single-user mode!

Today I have a master database that has gone slightly corrupt. The server
still runs, and dbcc reports that only one table is damaged (a messed up
page chain or something). I have a dump from yesterday, and I would like to
simply reload that dump into master. I have not made any recent changes in
master. In the past I would have simply used the three-step process to get
my database back.

My question is, can I use the old three-step method in ASE15, even though it
seems to have been removed from the manual? The new SAG seems to be saying
that when I start up in master-recover (-m) mode I will get some sort of
"generic" master database with a reset sa password, that I'll have to do
some stuff to before I can even load into it. [Does this have something to
do with the incorporation of buildmaster into dataserver?] I'm pretty sure
that in the old days when I started in single-user (-m) mode I'd see my
normal master database and could immediately load into it.

So, in my situation, when the master device is okay and all I want to do is
reload the master database, would it be true to say that there's now a
simple four-step process? Restart the database in master-recover mode
(with -m), update sysservers with my Backup Server info, issue a "load
database master" command, then restart the database in normal mode. It seems
like that should work, but I am nervous because it seems like something has
changed from what I'm accustomed to.

Another question: In older versions I'm pretty sure I could startup in -m
mode, then shutdown and startup without the -m, and my result would be just
exactly how I started. But the ASE15 System Administration Guide is
confusing, it implies that when you start in -m mode the master database
gets *overwritten* with some sort of generic master. Does that mean that if
you then restart without the -m, that your master database has been trashed?
The manual for dataserver doesn't make any mention of that, it just refers
to -m as "single-user" mode, like always. So what is with this "generic"
master that the SAG refers to?

Thanks.
- John.

[Mark A. Parsons]

Post by John Flynn
My question is, can I use the old three-step method in ASE15, even though it
seems to have been removed from the manual?

Yes.

Post by John Flynn
Another question: In older versions I'm pretty sure I could startup in -m
mode, then shutdown and startup without the -m, and my result would be just
exactly how I started. But the ASE15 System Administration Guide is
confusing, it implies that when you start in -m mode the master database
gets *overwritten* with some sort of generic master. Does that mean that if
you then restart without the -m, that your master database has been trashed?
The manual for dataserver doesn't make any mention of that, it just refers
to -m as "single-user" mode, like always. So what is with this "generic"

'-m' puts the dataserver into single-user mode, with the following message showing up in the ASE log:


*** WARNING ******************
ASE booted single user mode - updates allowed to system catalogs
*** WARNING ******************


If at this point you log in and load your master dump file you should find that the master database is loaded from said
dump file, the master database goes through the recovery process, the master database is brought online, and then the
dataserver is shutdown. IOW, same steps you're familiar with from old(er) versions of ASE.

--------------------------

As for why the ASE 15.x manuals have been changed/modified, or why certain parts are missing or (more) confusing ...
*shrug* ... this seems to be a common issue with ASE 15.x manuals.

I've got my own theories about the poor state of ASE documentation, but of course none of that matters nor will it 'fix'
the manuals.

I see two options:

1 - open a case with TechSupport; objective being to get questions answered *and* put in a CR to get the documentation
changed

or

2 - try out new (old?) commands on a test dataserver to see how they really work, ie, "When in doubt, try it out.";
obviously (?) maintain your own addendum to the doccumentation with your findings

[John Flynn]

Post by John Flynn
My question is, can I use the old three-step method in ASE15, even
though it seems to have been removed from the manual?

Yes.

OK, I did it. It worked perfectly, and exactly like it always has in
previous ASE versions. The good news is, my recovery of master seems to have
worked 100%.

I am mystified by what I see in the ASE15 manual. It says that in
master-recover (-m) mode there will be only one login name, namely sa, and
the sa password will be null. It also says sysservers will not be right
(assuming your BS is named anything other than SYB_BACKUP). But in reality,
when I restarted in -m mode, it did EXACTLY what I am accustomed to. My sa
password was intact, my sysservers were intact, my user databases were
present and useable, etc. I was able to reload my master database using the
exact same three-step sequence that I have used in prior versions.

Either my reading comprehension skills are going downhill, or the ASE15
manual is wack. Wack, I tells ya!

Thanks.
- John.

[Mark A. Parsons]

Post by John Flynn
Either my reading comprehension skills are going downhill

No ... at least not in *this* case. <g>

Post by John Flynn
the ASE15 manual is wack.

Yes ... unfortunately (especially for newbies). *sigh*

[Rob V [Sybase]]

Please submit your comments on the docs to ***@sybase.com.

Rob V.

Post by John Flynn

Post by John Flynn
My question is, can I use the old three-step method in ASE15, even
though it seems to have been removed from the manual?

Yes.

OK, I did it. It worked perfectly, and exactly like it always has in
previous ASE versions. The good news is, my recovery of master seems to
have worked 100%.
I am mystified by what I see in the ASE15 manual. It says that in
master-recover (-m) mode there will be only one login name, namely sa, and
the sa password will be null. It also says sysservers will not be right
(assuming your BS is named anything other than SYB_BACKUP). But in
reality, when I restarted in -m mode, it did EXACTLY what I am accustomed
to. My sa password was intact, my sysservers were intact, my user
databases were present and useable, etc. I was able to reload my master
database using the exact same three-step sequence that I have used in
prior versions.
Either my reading comprehension skills are going downhill, or the ASE15
manual is wack. Wack, I tells ya!
Thanks.
- John.

[Peter Collard]

Post by John Flynn
Either my reading comprehension skills are going downhill, or the ASE15
manual is wack. Wack, I tells ya!
Thanks.
- John.

Looks like they are describing a build from scratch with freshly rebuilt
master which could be what you are looking for if master is completely
stuffed.

Looks good to me!

[John Flynn]

Post by Peter Collard
Looks like they are describing a build from scratch with freshly
rebuilt master which could be what you are looking for if master is
completely stuffed.

I agree that what they're describing would seem in hindsight to apply to a
built-from-scratch scenario. However, if you read the actual text and try to
interpret it literally, it's difficult to read that meaning. Only if you
already know what the manual is trying to say would you be expected to read
it correctly. At best, it's extremely misleading.

Thanks.
- John.