V$archived_log and gv$archived_log

Question

2 node RAC on 10.2.0.4
Archived Logs stored in Local file systems of each node (not in ASM)
To determine the amount of redo generated on each nodes I googled and got the below query.
SELECT ROUND(SUM(blocks*block_size)/1024/1024/1024) arc_size,
TRUNC(first_time) arc_date
FROM v$archived_log
WHERE dest_id=1
GROUP BY TRUNC(first_time)
ORDER BY 2 DESCWhen i checked the actual size of archive logs on each nodes, i realized that the below query is returning the sum of archive logs of both nodes.(accurately though)
When i queried gv$archived_log, it returns
Size of archive logs from both nodes multiplied by 2 (inaccurate)Why can't v$archived_log just show the archivelog size of its own instance? Are there any other RAC related v$views which behaves similarly ?
Edited by: Herbaceous on May 26, 2011 8:13 AM

Answer 1

Hi,
Why can't v$archived_log just show the archivelog size of its own instance? Are there any other RAC related v$views which behaves similarly ?To understand It we need understand the difference between instance and database. An database (files) can be opened by many instances.
Database:
CONTROLFILE
DATAFILE
ONLINELOG
ARCHIVELOG
SPFILE
Instances:
PARAMETERS
MEMORY STRUCTURE
The fact that each instance have its own REDO/UNDO, but not mean they are part only of the instance (nodes).
So, archivelog doesn't have your own instance (node), but it's generate by one of instances and belong to database. So, it's recommended place archivelogs on a shared disk, such as ONLINELOGS.
You not need gv$ to query databases files, because it's the same on all instances.
gv$controlfile = (v$controlfile * qtd instances)
gv$datafile = (v$datafile * qtd instances)
gv$log = (v$log * qtd instances)
gv$archive_log = (v$archive_log * qtd instances)
So, I recommend you running your query specifying THREAD#, but the values can be inaccurate because THREAD# 2 can archive REDO of THREAD# 1.
Also you can use this query to estimate archivelogs size by THREAD.
SELECT distinct(to_char((bytes*0.000001),'9990.999')) size_mb FROM v$log;
  column ord     noprint
   column date_   heading 'Date'         format A15
   column no      heading '#Arch files'  format 9999999
   column no_size heading 'Size Mb'      format 9999999
   compute avg of no      on report
   compute avg of no_size on report
   break on report
   select MAX(first_time) ord, to_char(first_time,'DD-MON-YYYY') date_,
          count(recid) no, count(recid) *  <logfile_size> no_size
   from v$log_history
   group by to_char(first_time,'DD-MON-YYYY') ,THREAD#
   order by ord
   clear breaks
   clear computes
   clear columnsSee this tech note on MOS:
*Determine How Much Disk Space is Needed for the Archive Files [ID 122555.1]*
Regards,
Levi Pereira
Edited by: Levi Pereira on May 26, 2011 4:49 PM