OraFAQ Forum: Unix » zfs file system

Home » Infrastructure » Unix » zfs file system (12c)

Show: Today's Messages :: Polls :: Message Navigator
E-mail to friend

zfs file system [message #636484]

Sun, 26 April 2015 07:18

karthik4u
Messages: 36
Registered: September 2006
Location: CHENNAI

Member

My whole db is stored under '/oradb/TESTDB' which is nearly 350 GB is size.

I have zfs filesystem and I have allocated only one zfs pool for the file system '/oradb/TESTDB'.

Now please tell me whether creating two separate zfs pools for data files & redo log file path will increase the overall system performance.?

If so what are all the metrics I need to check before and after creating separate zfs pools?

Thanks

Report message to a moderator

Re: zfs file system [message #636485 is a reply to message #636484]

Sun, 26 April 2015 07:39

John Watson
Messages: 8976
Registered: January 2010
Location: Global Village

Senior Member

I can think of no reason why separate pools would increase performance. I would think that just one pool, striped across as many disks as possible, would be optimal. What do you think?

Report message to a moderator

Re: zfs file system [message #636486 is a reply to message #636485]

Sun, 26 April 2015 08:20

BlackSwan
Messages: 26766
Registered: January 2009
Location: SoCal

Senior Member

>Now please tell me whether creating two separate zfs pools for data files & redo log file path will increase the overall system performance.?

no change if you make more than 1 pool.

post SQL & results that show disk I/O is a system bottleneck.

Report message to a moderator

Re: zfs file system [message #636487 is a reply to message #636486]

Sun, 26 April 2015 08:32

karthik4u
Messages: 36
Registered: September 2006
Location: CHENNAI

Member

Find the below zfs details.

Code: [Select all] [Show/ hide]

$ zpool list
NAME    SIZE  ALLOC   FREE  CAP  HEALTH  ALTROOT
testpool    99G  34.4G  64.6G  34%  ONLINE  -
testpool1  572G   494G  78.1G  86%  ONLINE  -


$ zfs list
NAME                         USED  AVAIL  REFER  MOUNTPOINT
testpool                       83.3G  14.2G   106K  /testpool
testpool/ROOT                  30.0G  14.2G    31K  legacy
testpool/ROOT/s10s_u11wos      6.40G  23.6G  6.40G  /
testpool/crash                   31K  5.00G    31K  /crash
testpool/dump                  1.00G  14.2G  1.00G  -
testpool/home                  16.3G  3.71G  16.3G  /home
testpool/seos                   612M  2.40G   612M  /seos
testpool/swap                  8.25G  14.4G  8.00G  -
testpool/tmp                   2.11G  13.9G  2.11G  /tmp
testpool1                       541G  22.1G    31K  legacy
testpool1/TESTDB                156G  23.5G   156G  /oradb/TESTDB
testpool1/TESTDBDATA            262G  12.7G   262G  /oradb/TESTDB/TESTDBDATA
testpool1/appl1                   31K  1.80G    31K  /appl1
testpool1/appl2               74.8G  9.16G  74.8G  /appl2


$ zpool iostat -v
               capacity     operations    bandwidth
pool        alloc   free   read  write   read  write
----------  -----  -----  -----  -----  -----  -----
testpool       34.4G  64.6G      6      5   617K  55.8K
  c0d0s0    34.4G  64.6G      6      5   617K  55.8K
----------  -----  -----  -----  -----  -----  -----
testpool1       494G  77.7G    249    191  8.00M  9.73M
  c0d2s0     344G  54.3G    138     87  5.14M  5.77M
  c0d6       129G  20.0G     71     56  2.36M  3.20M
  c0d7      21.4G  3.44G     43     51   570K   856K
----------  -----  -----  -----  -----  -----  -----

$

Note that under /oradb/TESTDB only all data files and redo files are there.

Code: [Select all] [Show/ hide]

SQL> select * from v$logfile;

    GROUP# STATUS  TYPE    MEMBER                                                       IS_     CON_ID
---------- ------- ------- ------------------------------------------------------------ --- ----------
         3         ONLINE  /oradb/TESTDB/oradata/TESTDB/redo03.log                    NO           0
         2         ONLINE  /oradb/TESTDB/oradata/TESTDB/redo02.log                    NO           0
         1         ONLINE  /oradb/TESTDB/oradata/TESTDB/redo01.log                    NO           0
SQL>


SQL> select file_name from dba_data_files;

FILE_NAME
------------------------------------------------------------
/oradb/TESTDB/TESTDBDATA/TESTDBDATA07.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBDATA06.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBDATA05.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBDATA04.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBDATA03.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBDATA02.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBDATA.dbf

FILE_NAME
------------------------------------------------------------
/oradb/TESTDB/TESTDBDATA/TESTDBDATA01.dbf
/oradb/TESTDB/TESTDBDATA/TESTDBINDEX01.dbf
/oradb/TESTDB/TESTDBDATA/new_01.dbf

10 rows selected.

SQL>

Now tell how we can spilt?

Report message to a moderator

Re: zfs file system [message #636488 is a reply to message #636487]

Sun, 26 April 2015 08:37

BlackSwan
Messages: 26766
Registered: January 2009
Location: SoCal

Senior Member

>Now tell how we can spilt?

It appears that you are fixated on a solution in search of a problem.

post SQL & results that demonstrate a problem actually exits.

Report message to a moderator

Re: zfs file system [message #636489 is a reply to message #636488]

Sun, 26 April 2015 08:48

karthik4u
Messages: 36
Registered: September 2006
Location: CHENNAI

Member

Please provide me the sql commands to take during high IO activity related to zfs filesystem.

I will take and provide you the results.

Report message to a moderator

Re: zfs file system [message #636490 is a reply to message #636489]

Sun, 26 April 2015 08:55

BlackSwan
Messages: 26766
Registered: January 2009
Location: SoCal

Senior Member

unwilling or incapable to use GOOGLE yourself?

https://www.google.com/webhp?hl=en&tab=ww#hl=en&q=oracle+io+stats

Report message to a moderator

Re: zfs file system [message #636491 is a reply to message #636489]

Sun, 26 April 2015 10:59

John Watson
Messages: 8976
Registered: January 2010
Location: Global Village

Senior Member

If you want to calibrate the I/O rates on your hardware, use the ORION tool or dbms_resource_manager.caibrate_io. They are both documented in chapter 17 of the DB Performance Tuning Guide.

Report message to a moderator

Previous Topic:	Performance Bottle neck in Solaris
Next Topic:	/u01 file system size filling up

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

]

Current Time: Mon May 05 12:09:49 CDT 2025