Archive Logs

I am using DB2 8.2 and i've setup Archive logging on a database with the Active Log directory and Archive Log directory in different locations. I've also specified "use db2 to automatically archive the logs" and this is working fine.

What i'm wondering is when does DB2 actually decide that an "active" or "online archive" log is no longer required in the Active Log directory and moves it to the Archive Log directory. I know it does this when the database is stopped but the database i've setup it up on is active all the time and I haven't enabled any User Exit scripts.

Your logs get moved from the current logpath to the arhive log path whenever
the log file fills up, and this is irrespective of the state of any
transactions.
If you deactivate the Database normally, there will be no connections and all
transactions will be terminated.
The last acive log, also called the current active log is "chopped" and
archived.

If the archive log path becomes unaccessible, the logs are marked "to be
archived" and DB2 will wait until the archive path becomes available.
Meantime, the logs are kept in current log path.
HTH, Pierre.

--
Pierre Saint-Jacques
SES Consultants Inc.
514-737-4515
a écrit dans le message de news:
826461419.1136475464004.JavaMail.was... 7.sby.ibm.com...
>I am using DB2 8.2 and i've setup Archive logging on a database with the
>Active Log directory and Archive Log directory in different locations. I've
>also specified "use db2 to automatically archive the logs" and this is
>working fine.
>
> What i'm wondering is when does DB2 actually decide that an "active" or
> "online archive" log is no longer required in the Active Log directory and
> moves it to the Archive Log directory. I know it does this when the
> database is stopped but the database i've setup it up on is active all the
> time and I haven't enabled any User Exit scripts.

Hi Pete,

In 8.2 DB2 will archive the log either when it fills or when the last
connection to the database is closed, even if the log isn't full. If by
"active all the time" you mean there is always at least one connection to
the database, then you're only going to get logs archived when they fill.

Cheers,
Julian.

wrote in message
news:826461419.1136475464004.JavaMail.wassrvr-at-ltsg was007.sby.ibm.com...
>I am using DB2 8.2 and i've setup Archive logging on a database with the
>Active Log directory and Archive Log directory in different locations. I've
>also specified "use db2 to automatically archive the logs" and this is
>working fine.
>
> What i'm wondering is when does DB2 actually decide that an "active" or
> "online archive" log is no longer required in the Active Log directory and
> moves it to the Archive Log directory. I know it does this when the
> database is stopped but the database i've setup it up on is active all the
> time and I haven't enabled any User Exit scripts.

The other option is to periodically issue the "archive log" command which forces the closure of the "not full" active log.

Neil.

nntp://news.software.ibm.com/ibm.software.db2.udb/

Hi Pete,

In 8.2 DB2 will archive the log either when it fills or when the last
connection to the database is closed, even if the log isn't full. If by
"active all the time" you mean there is always at least one connection to
the database, then you're only going to get logs archived when they fill.

Cheers,
Julian.

wrote in message
news:826461419.1136475464004.JavaMail.wassrvr-at-ltsg was007.sby.ibm.com...
>I am using DB2 8.2 and i've setup Archive logging on a database with the
>Active Log directory and Archive Log directory in different locations. I've
>also specified "use db2 to automatically archive the logs" and this is
>working fine.
>
> What i'm wondering is when does DB2 actually decide that an "active" or
> "online archive" log is no longer required in the Active Log directory and
> moves it to the Archive Log directory. I know it does this when the
> database is stopped but the database i've setup it up on is active all the
> time and I haven't enabled any User Exit scripts.




[ibm.software.db2.udb]

Thanks for your help with this, it's what I thought happened but I couldn't find it mentioned anywhere.

Do you know how HADR works with Archive Logs? My question is to do with how HADR will re-sync the database if it's been offline for a couple of hours. Does it use the Archive Logs to perform this sync?

When an HADR standby is restarted, it will go through two distinct phases of
"catch up" processing.

1. Local catchup

In this phase, the standby restarts log replay. The restart location is
chosen much like it would be for a regular database restart / crash
recovery, i.e., far enough back so that (a) any updates that were only in
buffer pool cache and not yet on disk will be replayed over again, and (b)
meta data for the set of existing transactions can be rebuilt.

Any log files directly accessible to the HADR standby may be used in this
phase. This will include log files in the configured log paths on the
standby (whether left over from prior operation of the system or manually
placed there by the user) as well as any log files the standby can retrieve
from an archive (based on the log archiving configuration of the standby).
If the HADR pair happens to use a shared device (shared disk, storage
manager, etc.) it is possible that the standby will be able to directly
restore from archive log files that were generated on the primary while the
standby was offline.

After the HADR standby has processed all accessible log files in local
catchup phase, it will move into "remote catchup pending" state. At this
time the HADR standby will wait until it can establish a connection to the
HADR primary, if such a connection has not already been made during local
catchup phase.

2. Remote catchup

Once the HADR pair has a connection established, and after local catchup
completes on the HADR standby, the standby will request that log shipping
begin from its current log position. The HADR primary will now have the
responsibility to provide log data to the standby. In order to do so, the
HADR primary may access any log files available to it, either from the
configured log paths or by restoring log files from archive according to its
configuration. In this way, any log files that the HADR primary generated
while the standby was offline, and which the standby could not otherwise
access, will be read by the primary and shipped over the comm connection to
the standby. At the standby, the log data is written out to recreate the
log files and also replayed against the database, i.e., normal HADR standby
processing.

Once the HADR standby catches up to the current log position of the primary,
it will transition to Peer state where new log data is shipped from memory
as it is flushed on the primary, rather than being read back from existing
log files.

Regards,
- Steve P.
-----------------------------------------
Steve Pearson
IBM DB2 UDB for LUW Development
Portland, OR, USA

Thanks for the info, explains everything.