Subject | Re: [ib-support] Sweep vs Backup/Restore? |
---|---|
Author | Jason Chapman (JAC2) |
Post date | 2002-03-31T11:58:39Z |
Well the sweep is a little less intrusive, but I would would make the
following comments:
Backing up does a good garbage collection forward to the last rollback that
changed data.
Sweep does a good garbage collection forward to the oldest snapshot +
uncommited read committed data (I think).
One of the oldest problems we find are ghost connections or workstations
left on for ages with no activity. To clear these you need to stop the IB
services (of course when no one is using the DB).
Backup and restore (even if you don't make the restore live), proves data
integrity and highlights if there are any real problems in the DB, it is
what we use as it gives us peace of mind.
Frequency of the above reall depends on three things:
1) volatility of the data within the database
2) Load on the database server
3) Frequency of role backs during data changes.
I mention 3 as it has become practice for some of the developers I have
trained to make changes even if the probability of commiting is low
(something I think is not such a good idea).
A poorly groomed database that is small or has little load, doesn't need the
attention of constant sweeps etc, but with hundreds of users and chunky
queries having the best groomed database can be a huge advantage. So yet
again it is horse for courses.
Why don't you let us know a few of the metrics around your DB / Server /
Users and let us suggest a strategy?
Cheers,
Happy Easter (for those that celebrate it), and have good weekend to those
that don't.
JAC.
"GreatDayDan" <greatdaydan@...> wrote in message
news:20020330150335.66520.qmail@......
following comments:
Backing up does a good garbage collection forward to the last rollback that
changed data.
Sweep does a good garbage collection forward to the oldest snapshot +
uncommited read committed data (I think).
One of the oldest problems we find are ghost connections or workstations
left on for ages with no activity. To clear these you need to stop the IB
services (of course when no one is using the DB).
Backup and restore (even if you don't make the restore live), proves data
integrity and highlights if there are any real problems in the DB, it is
what we use as it gives us peace of mind.
Frequency of the above reall depends on three things:
1) volatility of the data within the database
2) Load on the database server
3) Frequency of role backs during data changes.
I mention 3 as it has become practice for some of the developers I have
trained to make changes even if the probability of commiting is low
(something I think is not such a good idea).
A poorly groomed database that is small or has little load, doesn't need the
attention of constant sweeps etc, but with hundreds of users and chunky
queries having the best groomed database can be a huge advantage. So yet
again it is horse for courses.
Why don't you let us know a few of the metrics around your DB / Server /
Users and let us suggest a strategy?
Cheers,
Happy Easter (for those that celebrate it), and have good weekend to those
that don't.
JAC.
"GreatDayDan" <greatdaydan@...> wrote in message
news:20020330150335.66520.qmail@......
> Good Morning!
>
> I have been reading the messages about Sweeps and
> was wondering: which is better; Sweep every night or
> B&R once a week?
>
> Thanks...Dan'l
>