"Log Shipping" Confusion

[r***@gmail.com]

We've talked about using log shipping for our recovery site for quite
a while. Here's the way we thought it would work:

- build recovery server, configured the same as primary
- setup database/transaction dump schedule
- have process on primary that looks for new dump files and copies
them to recovery node
- have process on recovery node that looks for new files copied
from primary.
- if new file is a full dump, load the database
- if new file is a log dump, apply the log

We were talking to some "in the know" folks last week and they said we
could run into trouble if the secondary server crashed, or was brought
down hard (no_wait). The explanation was that when the secondary
server was brought back up it would recover all the databases (as
expected). However, that recovery would increase the log sequence
number (not the correct term, I'm sure) by one. When we went to apply
the next log from the primary site, that log sequence number would
already have been applied, so the load would fail.

Is that true? Is this the correct way to maintain a recovery node
using log shipping?

TIA

[Richard]

Anybody?

[Sherlock, Kevin]

Hi Richard,
I don't know about anyone else, and maybe it's just my newsreader (IE6), but I
don't see the original posted question.

Can you re-post the question?

Post by Richard
Anybody?

[A. M.]

Post by Richard
Anybody?

You're original post hasn't shown up on Forums (where most of
us replying read them). Seems it might be a problem with Google
Groups. There have been spamming issues from Google Groups but
its unlikely to be related to your post not appearing.

Anyway, I went there to grab your original message.

Post by Richard
We've talked about using log shipping for our recovery site for quite
- build recovery server, configured the same as primary
- setup database/transaction dump schedule
- have process on primary that looks for new dump files and copies
them to recovery node
- have process on recovery node that looks for new files copied
from primary.

These two steps could be simplified into one. The routine
that does the dump of the primary connects to the secondary
and uses remote backup server to load the dumps. There's no
need to transfer the dump files this way and you don't need
processes scanning for files to load or copy.

Post by Richard
- if new file is a full dump, load the database
- if new file is a log dump, apply the log
We were talking to some "in the know" folks last week and they said we
could run into trouble if the secondary server crashed, or was brought
down hard (no_wait). The explanation was that when the secondary
server was brought back up it would recover all the databases (as
expected). However, that recovery would increase the log sequence
number (not the correct term, I'm sure) by one. When we went to apply
the next log from the primary site, that log sequence number would
already have been applied, so the load would fail.
Is that true? Is this the correct way to maintain a recovery node
using log shipping?

It may be true (I'm not sure what mechanism ASE uses) but its
not the correct way. Usual practice is to leave the database
offline when loading log dumps. Once you online it, you would
then need to load another db dump rather than a log dump.

I seem to recall Sybase made a change to this in recent versions
(you'll need to look up the details - I don't recall it right
now) but you don't mention which version you're using.

All in all, it might be simpler to use replication or ASE
Replicator.

-am © 2007

[Derek Asirvadem]

Post by r***@gmail.com
We were talking to some "in the know" folks last week and they said we
could run into trouble if the secondary server crashed, or was brought
down hard (no_wait). The explanation was that when the secondary
server was brought back up it would recover all the databases (as
expected). However, that recovery would increase the log sequence
number (not the correct term, I'm sure) by one. When we went to apply
the next log from the primary site, that log sequence number would
already have been applied, so the load would fail.
Is that true? Is this the correct way to maintain a recovery node
using log shipping?

(Thanks AM for posting the original)

No it isn't. Your folks "in the know" do not appear to have actual
experience with the subject on Sybase and may be speculating. Log
shipping works just fine for many, many years. Replication Server is
expensive if you are going to use it just to ship log dumps.

AM is right, there are changes (improvements) in 12.x, white papers,
PDFs you can download. The old method works, the new options provide a
bit more flexibility. Look up "... {for|with} standby_access"

These days many people are keeping the Recovery site Db = Offline,
while feeding it tran dumps; if you want to make it available, just
"online database". Some people are using the Recovery site Db =
Online, offloading some of their reporting load onto it, which is a bit
more complicated (ensure you have the licencing for it and it easy).

[Richard]

Thanks for the replies.

We are currently running 12.5.0.3 but we are upgrading to 15.0 within
six months.

Richard