How to make sure that the data in stby is not more than 15 minutes

Hi
I have a standby machine. I am asked to make sure that, in case of failure, The standby is not older than 15 minutes. As per the transactions, the new archive is getting created every 15 to 60 minutes. And as soon as the new Archive is created, then it is getting transferred to Standby machine. What if log swicth does not occur for say 40 minutes. And at 20th minute the system failed. And I cannot make sure that the data is just older of 15minutes in Standby.

My question is, is there any way I can make sure that, the STANDBY is just not older than 15 minutes.

just suggestion, can you set up a batch job, wake up every 15 min check if no arch file generated, issue
alter system switch logfile;

That's what I am doing. My batch wakes up every fifteen minutes and if any new Archive logfile occurs, it transfers it to Standby. My question is not that.

What if there is not Logwitch for 60 minutes. Then Production failes, and the data in Online redo logs are not transferred for the past 60 minutes, and I will loose the data. How to make sure that, I will not loose data not older than 15 minutes

how about set log_time_interval=900

you batch job can do:

if [ -a /ora/arch/xxxx ] ; then
svrmgrl < connect internal
alter system switch logfile;
EXIT
end_sql
fi

or use find -atime .... execute force_switch.sh ... {}

Hi Pando and Jm

Thanks. I had that in mind forcing the logswitch for evry 15 minutes / 900 sec. But to make sure with you experts .

Is there any other idea to do it or just follow the log switch procedure.

How about check_point_timeout ?

Guys,

Even I am eager to know that. I was above to suggest the same 'switching log file for evry 15 min'. But, didn't feel its the best way of doing it... so, I didn't post it and waiting for someone to come up with best way of doin it..

Curious, how is alter switch log different from log_time_interval=XXX

I make a pardon, the interval is based on OS BLOCKS, I really meant timeout that´s why I set it to 900 :P
Sorry for the confusion!

These two parms(CHECKPOINT and TIME) are different. One specified in seconds and other in # of blocks.

LOG_CHECKPOINT_INTERVAL specifies the frequency of checkpoints in terms of the number of redolog file blocks(OS blocks, not DB blocks) that can exist between an incremental checkpoint and the last block written to the redo log.

Regardless of this value, a checkpoint always occurs when switching from one online redo log file to another. Therefore, if the value exceeds the actual redo log file size, checkpoints occur only when switching logs.

Analyze this, assuming 512k is redolog size and 520k is LOG_CHECKPOINT_INTERVAL( interms of blocks). Though you specify at this INTERVAL checkpoint to occur, it occurs only when log switch occurs, as its just 512k.

Hope this helps...

[Edited by sreddy on 12-29-2000 at 11:20 PM]

Thanks all, especially Sreddy and Pando. I got it. Giving right parameters for Log Interval and Time out ensures the # of minutes desired recovery.

Thanks Everyone and I wish Great New Year for the whole Forum Group.

Quote:

---

Originally posted by sreddy
I got it. Do what Pando suggested.

LOG_CHECKPOINT_TIMEOUT = 900

It will force log switch to occur as it flushes the dirty buffers(in cache) to redolog after specified time interval in sec.

---

Unfortunately this wil not force a log switch. It will force a checkpoint everey 900 seconds, but the checkpoint and the log switch are not the same thing.

There is no init parameter to force a periodic log switch. The only way to do it is through a batch job, preferably through DBMS_JOB.

indeed, checkpoint wont cause log switch may be as suggested in previous try dbms_job

Yep! The only alternative way(other than using log switch command for every 15 min) is you can make log switch to occur :

Resize your logs in such a way log switch occurs every 15 minute until you upgrade your database to 9i which is not gonna happen soon.

This z what I have come up with at last, which you know already! :)


[Edited by sreddy on 01-02-2001 at 04:57 PM]

Hi SReddy,

We cannot always make sure to have a log switch by resizing the log file. I think it is very difficult if not impossible to plan a log file which fills every 15 minutes on production machine. As some times when they are doing some processing it gets filled every 1 minute. And some times it will not get filled even for 1 hour.

So is the reason I am looking for a better solution

Sizing your logfiles to switch every x minutes is next to impossible in the real world. How fast a logfile fills up is directly proportional to the transaction volume. If you size your logfiles to swith every x minutes during the heaviest transaction load, they will be oversized for normal load and might not even switch at other times. If you size your logfiles for normal load, they may switch quickly during peak periods. The key is to strike a balance between recoverability and performance with the business will dictate.


Kishore has a reasonable request. "How do I make sure my database does a logswitch every x minutes?" As several have suggested, it can only be done via a scheduled job (cron, batch, dbms_job, etc.) There are currently no init.ora parameters that will periodically force a log switch.

Thats TRUE, we can't predict the transaction volume at any point of time. So, we end up having no alternative to switch log file at specified time interval other than issuing a command. We got to go with that for 6 months more until they release 9i and it becomes stable... Good thread! Thanks for participants...

[Edited by sreddy on 01-02-2001 at 06:03 PM]

waitingfor 9i to become stable until end of this year .... probably :D