RAM issue RAC

[Roshan]

Oracle DB 12.1.0.2 (2 node)
Solaris 11.4

Hello Erman,

Do you know why first node is using more RAM than second node? When DB is idle, I noticed the following connections where present. These were mostly Golden Gate connections (downstream replication).

dware1 is connection string using service name. ORAGGEXT is Golden Gate extract user

293 53724 ORAGGEXT dware1 SYS 11/11/2019 23:46:08 oracle@dware1 (AS06) ora12c ACTIVE 23523

572 31201 ORAGGEXT dware1 ORAGGEXT 22/10/2019 20:34:15 extract@dware1 (TNS V1-V3) ora12c ACTIVE 12011

575 58361 BIREPORT dware1 BIREPORT 12/11/2019 00:08:00 sqlplus@dware1 (TNS V1-V3) ora12c ACTIVE 3976

855 38798 SYS dware1 SYS 22/10/2019 20:34:16 oracle@dware1 (MS00) ora12c ACTIVE 12379

865 17910 ORAGGEXT dware1 SYS 11/11/2019 21:34:15 oracle@dware1 (AS0L) ora12c ACTIVE 19399

1138 58983 SYS dware1 SYS 22/10/2019 20:34:16 oracle@dware1 (MS01) ora12c ACTIVE 12381

1424 7172 SYS dware1 SYS 22/10/2019 20:34:16 oracle@dware1 (MS02) ora12c ACTIVE 12383

1705 25739 SYS dware1 SYS 22/10/2019 20:34:16 oracle@dware1 (MS03) ora12c ACTIVE 12385

1996 12882 SYS MTG\RHIS-DBA-VPN SYS 12/11/2019 00:07:26 Toad.exe roshanbi ACTIVE 12340:5756

2272 34827 ORAGGEXT dware1 SYS 11/11/2019 23:16:28 oracle@dware1 (AS08) ora12c ACTIVE 9675

2282 64038 ORAGGEXT dware1 SYS 11/11/2019 23:34:51 oracle@dware1 (AS0Q) ora12c ACTIVE 18691

2552 2605 ORAGGEXT dware1 SYS 11/11/2019 23:24:40 oracle@dware1 (AS0E) ora12c ACTIVE 13526

2557 19828 ORAGGEXT dware1 SYS 11/11/2019 22:04:47 oracle@dware1 (AS0J) ora12c ACTIVE 3242

2567 1095 ORAGGEXT dware1 SYS 11/11/2019 20:07:35 oracle@dware1 (AS0K) ora12c ACTIVE 28898

2836 18600 SYS dware1 SYS 29/10/2019 20:13:57 oracle@dware1 (AP02) ora12c ACTIVE 1020

3405 49607 ORAGGEXT dware1 SYS 11/11/2019 23:36:45 oracle@dware1 (AS0A) ora12c ACTIVE 19249

3694 15871 ORAGGEXT dware1 SYS 29/10/2019 20:13:58 oracle@dware1 (AS01) ora12c ACTIVE 1022

3969 8526 ORAGGEXT dware1 SYS 12/11/2019 00:01:03 oracle@dware1 (AS0B) ora12c ACTIVE 734

Is it because cluster is Active/Passive?

Regards,

Roshan

[ErmanArslansOracleBlog]

You are using RAC, right? You have at least 2 instances(up and running) and a service name which is actively being listened by both the first instance and second instance right? You have 2 local listeners, 1 for each of your instance and 3 scan listeners right?
"first node is using more RAM than second node?" -> how did you measure that?
Note that, this is normally possible.. (if you are measuring this by taking the process/PGA memory into your consideration in addition to SGA.)

How did you configured your client, app tier and goldengate connections? If you want a good load balancing, you need to use a scan-name-based + servicename-based connection string..

[Roshan]

Load balancing is configured for the RAC.

I see some memory warnings in EM 12c.

streams_pool.JPG

What are the signs to tell RAM needs to be increased?

[ErmanArslansOracleBlog]

Then there must be something,that uses steams in the background, is running and connected on the first node of your RAC database.
The Streams pool is a shared memory pool. Is is the amount of memory a process can use from the Streams pool.

Oracle GoldenGate, XStream, Oracle Streams, Oracle Advanced Queuing, and Oracle Data Pump use e Streams pool.. There can be other products (+ customized ones) and features that use Streams as well.
So any load on these types of products may affect your stream pool usage.
Check your db accordingly.

[Roshan]

I am using Oracle Golden Gate on RAC.

GGSCI (dware1) 1> fo all

Program     Status      Group       Lag at Chkpt  Time Since Chkpt

MANAGER     RUNNING
EXTRACT     RUNNING     EXTDWH      unknown       00:00:09
EXTRACT     RUNNING     PXDWN       00:00:00      00:00:06
REPLICAT    RUNNING     REPLOGGT    00:00:00      00:00:08

The pump process is running with first node IP. It is not accepting the scan host name.

Pump trail file:

EXTRACT PXDWN
USERIDALIAS oraggext
RMTHOST 10.215.22.99, MGRPORT 7819
RMTTRAIL /ogg/ora12c/dirdat/cd
PASSTHRU
TABLE CBS_CORE.TEST;

[ErmanArslansOracleBlog]

Okay.. Yes it seems  so..
Also, then this may explain that weirdness in streams pool size.

Did you do a proper sizing when configuring this env?


The general formula for calculating the minimum streams_pool_size is as follows;

(<number of extracts> * <Extract memory size in GB))
+
(<number of extracts>* <Extract memory size in GB>*0.25)

Also read ->  https://www.ateam-oracle.com/memory-requirement-for-oracle-goldengate-integrated-extract