| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 5015.1 | | NOVA::R_ANDERSON | Oracle Corporation (603) 881-1935 | Tue Feb 11 1997 09:09 | 9 |
| Your customer is probably applying the AIJ journals out of sequence.
This situation is not allowed in Rdb7 but is "tolerated" in Rdb v6.1 and earlier
because the customer is allowed to try to apply any journal regardless of
whether it can be applied.
Recommendation: Start the recovery with the previous AIJ journal.
Rick
|
| 5015.2 | aborts with the aij it says it needs | UKVMS3::SHISCOCK | stand and deliver | Tue Feb 11 1997 09:38 | 12 |
|
The database backup was created on-line also. Once restored it
says it needs aij 136 which is the one that bugchecks. So
he's trying to apply 136 and 137.
He claims to have tried 135,136 & 137 but says that 135 is too
old. Even so I can't see with two journals he could get them
out of sequence. Wouldn't it complain about the sequence numbers
in the first place.
thanks so far,
Steve
|
| 5015.3 | Looks strange | svrav1.au.oracle.com::MBRADLEY | I was dropped on my head as a baby. What's your excuse? | Tue Feb 11 1997 18:53 | 9 |
| Rdb 6.1 and earlier weren't that good at handling noquiet point AIJ
backups.
Having said that, I am surprised to see an illegal page count error from
doing this.
G'day,
Mark.
|
| 5015.4 | -<Looks familiar...>-< Looks familiar... >-< Looks familiar...>-< Looks familiar...> | NOVA::BALL_A | | Wed Feb 12 1997 14:02 | 5 |
| The ILLPAGCNT is being returned from a call to $expreg requesting 33630
pagelets - we are going to take a look at the AIJ itself - and Hi Mark,
this looks awfully like the 'post scenario...
ALan
|
| 5015.5 | possibly the aij, but it may be the db as well | NOVA::BRYDEN | | Wed Feb 12 1997 14:29 | 12 |
| Alan,
The customer is using noquiet point for both aij and
database? When they restored teh database backup, which backup was
it? The database needs to be restored to a known good point and
then incrementals rolled forward.
Does the customer have a quietpoint backup of their
database that they can go back to and then roll forward? As you
said what were teh first few blocks in the aij?
Dave
|
| 5015.6 | okay on quiet-point test | UKVMS3::SHISCOCK | stand and deliver | Thu Feb 13 1997 03:27 | 7 |
|
The customer did a new test using a quiet-point db backup and
this worked. Guess they'll need to re-think using noquiet
aij and db backups.
cheers,
Steve
|
| 5015.7 | | NOVA::R_ANDERSON | Oracle Corporation (603) 881-1935 | Thu Feb 13 1997 07:43 | 4 |
| ...at least until the next ECO comes out, which includes enhancements to better
detect this problem...
Rick
|
| 5015.8 | do quiet on weekend, noquiet during week.. | NOVA::BRYDEN | | Thu Feb 13 1997 14:19 | 18 |
| Steve,
We had exactly the same thing happen down here a while
back.... Custoemr took a quiet point backup and then subsequent
backups were no-quietpoint.... day 65 comes along and they have a
failure, tried the latest noquietpoint backup, restored and rolled
forward the aij.... it failed. They ultimately had to go back to
the quietpoint from 65 days earlier and then luckily roll forward
all teh aij's.
What this site did was a full quietpoint on the weekends
and then noquietpoint incrementals during the week. this meant
they only had to go back a max of 6 days to get a synch point.
Were you able to determine whether it was the aijs, that
caused the cproblem or was it the state of teh db?
Dave
|
| 5015.9 | We should see from the dump | svrav1.au.oracle.com::MBRADLEY | I was dropped on my head as a baby. What's your excuse? | Sun Feb 16 1997 19:08 | 13 |
| As has been said, a dump of the first few blocks of the AIJ file would be
helpfull.
In any event, even with the next ECO of 6.1, which may not evidence this
problem, you still may have to go baclk to the most recent noqiet point AIJ
backup to get a good roll forward.
In V7 you only have to go back to the most recent of a quiet database or a
quiet AIJ backup.
G'day,
Mark.
|