Issue installing the 11gR2 database software after the 12.1.0.2 clusterware setup

If you don’t plan to install the 11gR2 database software after a 12.1.0.2 clusterware installation, I guess there is no need for you to read this post.

I just want to share the issue I got and the way you could workaround it. The purpose of this post is just to save your time, in case of.

So, after a 12.1.0.2 clusterware installation (on a 2 nodes RAC cluster), I decided to install the 11.2.0.4 database software. I launched the runInstaller, followed the install process until the Step 4:

As you can see the nodes that are part of the cluster have not been proposed and clicking next would produce “CRS is not installed on any of the nodes”.

The first question you may ask is: Is the 11gR2 database compatible with CRS 12.1? Yes it is, as you can see here:

As I am curious, I canceled the 11gR2 database software installation and gave a try with the 12.1.0.2 database software. The Step 4 produced:

As you can see the nodes that are part of the cluster have been proposed.

I canceled the 12.1.0.2 database software installation and I had a look to the Inventory. Then I discovered that the CRS=”true” flag was not set for the 12.1.0.2 GI HOME:

$ cat /etc/oraInst.loc | grep inventory_loc
inventory_loc=/ec/poc/server/oracle/olrpoc1/u000/oraInventory

$ cat /ec/poc/server/oracle/olrpoc1/u000/oraInventory/ContentsXML/inventory.xml
..
HOME NAME="OraGI12Home1" LOC="/ec/poc/server/oracle/olrpoc1/u000/product/GRID.12.1.0.2" TYPE="O" IDX="1"
..

Then I added the the CRS=”true” flag that way:

$GI_HOME/oui/bin/runInstaller -updateNodeList ORACLE_HOME="/ec/poc/server/oracle/olrpoc1/u000/product/GRID.12.1.0.2" CRS=true

So that:

$ cat /ec/poc/server/oracle/olrpoc1/u000/oraInventory/ContentsXML/inventory.xml
..
HOME NAME="OraGI12Home1" LOC="/ec/poc/server/oracle/olrpoc1/u000/product/GRID.12.1.0.2" TYPE="O" IDX="1" CRS="true"
..

Then I relaunched the 11.2.0.4 database software installation, and the Step 4 produced:

As you can see the nodes have been proposed so that I have been able to complete the installation successfully.

Remarks:

MOS 1053393.1 provides more details about the CRS=”true” flag.
I guess the “issue” will be the same for database software >=10.1 and < 12.1 (But I did not test it).
On another RAC, I upgraded the GI from 12.1.0.1 to 12.1.0.2 and then the CRS=”true” flag has been set automatically for the 12.1.0.2 GI.

Conclusion: After a 12.1.0.2 CRS installation,

You may need to put the CRS=”true” flag to avoid this issue during a 11gR2 database software installation.
You don’t need to put the CRS=”true” flag during a 12.1.0.2 database software installation as the issue does not appear.
It looks like you won’t hit the issue if the GI has been upgraded to 12.1.0.2 (See third remark).

ASM Rebalance: Why is the avg By/Read equal to 1MB while the allocation unit is 4MB?

During the “A closer look at ASM rebalance series” I observed that the average By/Read extracted from the v$asm_disk_stat view during a rebalance is equal to 1MB even if the ASM allocation unit is set to 4MB.

Example:

The question is: why?

To try to answer this question I decided to use strace (the Linux system call trace utility) on the ASM arb process (this process rebalances data extents within an ASM disk group) during the rebalance of the DATA diskgroup (AU size is 4MB).

From strace I can see:

That a very large majority (more than 80% in my case) of the IOs are submitted that way:

io_submit(140691479359488, 4, {{0x7ff547f6c210, 0, 0, 0, 261}, {0x7ff547f6d0b0, 0, 0, 0, 261}, {0x7ff547f6c960, 0, 0, 0, 261}, {0x7ff547f6cbd0, 0, 0, 0, 261}}) = 4

We can see that 4 IOs have been submitted at the same time (with a single io_submit call).

And that about 100% of all the IOs submitted are 1MB (look at the IO size which is 1048576 bytes):

io_getevents(140691479359488, 1, 128, {{0x7ff547f6c960, 0x7ff547f6c960, 1048576, 0}, {0x7ff547f6cbd0, 0x7ff547f6cbd0, 1048576, 0}, {0x7ff547f6d0b0, 0x7ff547f6d0b0, 1048576, 0}, {0x7ff547f6c210, 0x7ff547f6c210, 1048576, 0}}, {600, 0}) = 4

I also straced arb during the rebalance of the DATA1M diskgroup (Allocation unit of 1MB) and observed:

That about 80% of the IOs are submitted that way:

io_submit(139928633700352, 1, {{0x7f43aadbd210, 0, 1, 0, 262}}) = 1

So 1 IO is submitted per io_submit call.

And that about 100% of all the IOs submitted are 1MB:

io_getevents(139928633700352, 3, 128, {{0x7f43aadbd6f0, 0x7f43aadbd6f0, 1048576, 0}}, {0, 0}) = 1

So that it makes sense to conclude that:

The arb process always request IOs of 1MB (whatever the allocation unit is).
The arb IOs requests are grouped and submitted at the same time depending of the diskgroup allocation unit (4 IOs are submitted at the same time with allocation unit set to 4MB, 1 IO submitted by io_submit call with allocation unit set to 1MB).

Based on this, I think it makes sense that the v$asm_disk_stat view reports Avg By/Read of 1MB during the rebalance process (whatever the allocation unit is).

But now one more question: What is the impact of the Linux Maximum IO size?

For this, I launched the rebalance 2 times (on both diskgroups: DATA with AU=4MB and DATA1M with AU=1MB): One time with max_sectors_kb set to 512 and one time with max_sectors_kb set to 4096.

And observed that:

Oracle still does 1MB IOs only and groups the IOs depending of the allocation unit (whatever the max_sectors_kb is).

With max_sectors_kb set to 512: The IOs issued on the devices linked to the ASM disks are limited to 512 KB (even if oracle requested 1MB IOs). This can be seen with the iostat output (looking at the avgrq-sz field):

Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util
sdak              0.00     0.00    0.00   13.65     0.00 13919.15  1019.72     0.04    3.11   3.11   4.25
sdbd              0.00     0.00    0.10    0.00   102.05     0.00  1020.50     0.00    2.00   2.00   0.02
sddh              0.00     0.00    0.15    0.00   153.25     0.00  1021.67     0.00    4.33   4.33   0.07
sdej              0.00     0.00    0.00   12.35     0.00 12640.10  1023.49     0.04    3.01   3.01   3.72
sdcr              0.00     0.00    0.00   11.80     0.00 12083.20  1024.00     0.03    2.71   2.71   3.20
sdfi              0.00     0.00    0.00   12.35     0.00 12646.40  1024.00     0.07    5.55   5.55   6.85
sdg               0.00     0.00    0.00   11.70     0.00 11980.80  1024.00     0.06    5.35   5.34   6.25

The avgrq-sz is the average size (in sectors) of the requests that were issued to the device. The sector size I am using is 512 bytes (not 4K) so that the IOs are about 1024*512 bytes = 512 KB (which is our max_sectors_kb size).

With max_sectors_kb set to 4096: The IOs issued on the devices linked to the ASM disks can be greater than 1MB (even if oracle requested 1MB IOs):

Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util
sdi               0.00     0.00    3.75    2.05 20173.15  2253.60  3866.68     0.17   29.72  28.79  16.70
sdas              0.00     0.00    3.90    1.95 20992.35  1638.80  3868.57     0.18   31.08  29.91  17.50
sdgj              0.00     0.00    3.75    2.00 20173.85  2253.20  3900.36     0.20   34.71  33.74  19.40

about 2MB in my case (3900 * 512 bytes).

Remarks:

It has been tested with ASM 11.2.0.4 on Linux x86-64 (without asmlib).
I’ll update this post with ASM 12c results as soon as I can (if something new needs to be told).

Conclusion:

The arb process always request IOs of 1MB (whatever the allocation unit is).
The arb process always request IOs of 1MB (whatever the max_sectors_kb is): Then it looks like arb doesn’t probe the IO capabilities of the associated device.
The arb IOs requests are grouped and submitted at the same time depending of the diskgroup allocation unit.
The kernel splits the arb IOs requests if max_sectors_kb is < 1MB.
The kernel try to merge the arb IOs requests if max_sectors_kb is > 1MB.

Thanks to Frits Hoogland for his “Extra huge database IOs series” and for the time he spent answering the questions I asked.

A closer look at ASM rebalance, Part III: Disks have been added and dropped (at the same time)

This article is the third Part of the “A closer look at ASM rebalance” series:

Part I: Disks have been added.
Part II: Disks have been dropped.
Part III: Disks have been added and dropped (at the same time).

If you are not familiar with ASM rebalance I would suggest first to read those 2 blog posts written by Bane Radulovic:

In this part III I want to visualize the rebalance operation (with 3 power values: 2,6 and 11) after disks have been added and dropped (at the same time).

To do so, on a 2 nodes Extended Rac Cluster (11.2.0.4), I added 2 disks and dropped 2 disks (with a single command) into the DATA diskgroup (created with an ASM Allocation Unit of 4MB) and launched (connected on +ASM1):

alter diskgroup DATA rebalance power 2; (At 02:11 PM).
alter diskgroup DATA rebalance power 6; (At 02:24 PM).
alter diskgroup DATA rebalance power 11; (At 02:34 PM).

And then I waited until it finished (means v$asm_operation returns no rows for the DATA diskgroup).

Note that 2) and 3) interrupted the rebalance in progress and launched a new one with a new power.

During this amount of time I collected the ASM performance metrics that way for the DATA diskgroup only.

I’ll present the results with Tableau (For each Graph I’ll keep the “columns”, “rows” and “marks” shelf into the print screen so that you can reproduce).

Note: There is no database activity on the Host where the rebalance has been launched.

Here are the results:

First let’s verify that the whole rebalance activity has been done on the +ASM1 instance (As I launched the rebalance operations from it).

We can see:

That all Read and Write rebalance activity has been done on +ASM1 .
That the read throughput is very close to the write throughput on +ASM1.
The impact of the power values (2,6 and 11) on the throughput.

Now I would like to compare the behavior of 3 Sets of Disks: The disks that have been dropped, the disks that have been added and the other existing disks into the DATA diskgroup.

To do so, let’s create in Tableau 3 groups:

Let’s call it “3 Groups”

So that now we are able to display the ASM metrics for those 3 sets of disks.

I will filter the metrics on ASM1 only (to avoid any “little parasites” coming from ASM2).

Let’s visualize the Reads/s and Writes/s metrics:

We can see that during the 3 rebalances:

No writes on the dropped disks.
No reads on the new disks.
Number of Reads/s increasing on the dropped disks depending of the power values.
Number of Writes/s increasing on the new disks depending of the power values.
Reads/s and Writes/s both increasing on the other disks depending of the power values.
As of 03:06 PM, no activity on the dropped and new disks while there is still activity on the other disks.

Are 1, 2, 3, 4 and 5 surprising? No.
What happened for 6? I’ll answer later on.

Let’s visualize the Kby Read/s and Kby Write/s metrics:

We can see that during the 3 rebalances:

No Kby Write/s on the dropped disks.
No Kby Read/s on the new disks.
Number of Kby Read/s increasing on the dropped disks depending of the power values.
Number of Kby Write/s increasing on the new disks depending of the power values.
Kby Read/s and Kby Write/s both increasing on the other disks depending of the power values.
Kby Read/s and Kby Write/s are very close on the other disks (It was not the case into the Part I).
As of 03:06 PM, no activity on the dropped and new disks while there is still activity on the other disks.

Are 1, 2, 3, 4, 5 and 6 surprising? No.
What happened for 7? I’ll answer later on.

Let’s visualize the Average By/Read and Average By/Write metrics:

Important remark regarding the averages computation/display: The By/Read and By/Write measures depend on the number of reads. So the averages have to be calculated using Weighted Averages.

Let’s create the calculated field in Tableau for the By/Read Weighted Average:

The same has to be done for the By/Write Weighted Average.

Let’s see the result:

We can see:

The Avg By/Read on the dropped disks is about the same (about 1MB) whatever the power value is.
The Avg By/Write on the new disks is about the same (about 1MB) whatever the power value is.
The Avg By/Read and Avg By/Write on the other disks is about the same (about 1MB) whatever the power value is.

Are 1,2 and 3 surprising? No for the behaviour,Yes (at least for me) for the 1MB value as the ASM allocation unit is 4MB.

Now that we have seen all those metrics, we can ask:

Q1: So what the hell happened at 03:06 pm?

Let’s check the alert_+ASM1.log file at that time:

Mon Aug 25 15:06:13 2014
NOTE: membership refresh pending for group 4/0x1e089b59 (DATA)
GMON querying group 4 at 396 for pid 18, osid 67864
GMON querying group 4 at 397 for pid 18, osid 67864
NOTE: Disk DATA_0006 in mode 0x0 marked for de-assignment
NOTE: Disk DATA_0007 in mode 0x0 marked for de-assignment
SUCCESS: refreshed membership for 4/0x1e089b59 (DATA)
NOTE: Attempting voting file refresh on diskgroup DATA
NOTE: Refresh completed on diskgroup DATA. No voting file found.
Mon Aug 25 15:07:16 2014
NOTE: stopping process ARB0
SUCCESS: rebalance completed for group 4/0x1e089b59 (DATA)

We can see that the ASM rebalance started the compacting phase (See Bane Radulovic’s blog post for more details about the ASM rebalances phases).

Q2: The ASM Allocation Unit size is 4MB and the Avg By/Read is stucked to 1MB,why?

I don’t have the answer yet, it will be the subject of another post.

Two remarks before to conclude:

The ASM rebalance activity is not recorded into the v$asm_disk_iostat view. It is recorded into the v$asm_disk_stat view. So, if you are using the asm_metrics utility, you have to change the asm_feature_version variable to a value > your ASM instance version.
I tested with compatible.asm set to 10.1 and 11.2.0.2 and observed the same behaviour for all those metrics.

Conclusion of Part III:

Nothing surprising except (at least for me) that the Avg By/Read is stucked to 1MB (While the allocation unit is 4MB).
We visualized that the compacting phase of the rebalance operation generates much more activity on the other disks compare to near zero activity on the dropped and new disks.
I’ll update this post with ASM 12c results as soon as I can (if something new needs to be told).

A closer look at ASM rebalance, Part II: Disks have been dropped

This article is the second Part of the “A closer look at ASM rebalance” series:

Part I: Disks have been added.
Part II: Disks have been dropped.
Part III: Disks have been added and dropped (at the same time).

If you are not familiar with ASM rebalance I would suggest first to read those 2 blog posts written by Bane Radulovic:

In this part II I want to visualize the rebalance operation (with 3 power values: 2,6 and 11) after disks have been dropped.

To do so, on a 2 nodes Extended Rac Cluster (11.2.0.4), I dropped 2 disks into the DATA diskgroup (created with an ASM Allocation Unit of 4MB) and launched (connected on +ASM1):

alter diskgroup DATA rebalance power 2; (At 09:09 AM).
alter diskgroup DATA rebalance power 6; (At 09:19 AM).
alter diskgroup DATA rebalance power 11; (At 09:29 AM).

And then I waited until it finished (means v$asm_operation returns no rows for the DATA diskgroup).

Note that 2) and 3) interrupted the rebalance in progress and launched a new one with a new power.

During this amount of time I collected the ASM performance metrics that way for the DATA diskgroup only.

I’ll present the results with Tableau (For each Graph I’ll keep the “columns”, “rows” and “marks” shelf into the print screen so that you can reproduce).

Note: There is no database activity on the Host where the rebalance has been launched.

Here are the results:

First let’s verify that the whole rebalance activity has been done on the +ASM1 instance (As I launched the rebalance operations from it).

We can see:

That all Read and Write rebalance activity has been done on +ASM1 .
That the read throughput is very close to the write throughput on +ASM1.
The impact of the power values (2,6 and 11) on the throughput.

Now I would like to compare the behavior of 2 Sets of Disks: The disks that have been dropped and the disks that will remain into the DATA diskgroup.

To do so, let’s create in Tableau a SET that contains the 2 dropped disks.

Let’s call it “Dropped Disks”

So that now we are able to display the ASM metrics IN this set (the 2 dropped disks) and OUT this set (the disks that will remain into the DATA diskgroup).

I will filter the metrics on ASM1 only (to avoid any “little parasites” coming from ASM2).

Let’s visualize the Reads/s and Writes/s metrics:

We can see that during the 3 rebalances:

No writes on the dropped disks.
Number of Reads/s increasing on the dropped disks depending of the power values.
Reads/s and Writes/s both increasing on the remaining disks depending of the power values.

Are 1, 2 and 3 surprising? No.

Let’s visualize the Kby Read/s and Kby Write/s metrics:

We can see that during the 3 rebalances:

No Kby Write/s on the dropped disks.
Number of Kby Read/s increasing on the dropped disks depending of the power values.
Kby Read/s and Kby Write/s both increasing on the remaining disks depending of the power values.

Are 1, 2 and 3 surprising? No.

Let’s visualize the Average By/Read and Average By/Write metrics:

Let’s create the calculated field in Tableau for the By/Read Weighted Average:

The same has to be done for the By/Write Weighted Average.

Let’s see the result:

We can see:

The Avg By/Read on the dropped disks is about the same (about 1MB) whatever the power value is.
The Avg By/Read and Avg By/Write on the remaining disks is about the same (about 1MB) whatever the power value is.

Are 1 and 2 surprising? No for the behaviour, Yes (at least for me) for the 1MB value as the ASM allocation unit is 4MB.

Now that we have seen all those metrics, we can ask:

Q1: The ASM Allocation Unit size is 4MB and the Avg By/Read is stucked to 1MB,why?

I don’t have the answer yet, it will be the subject of another post.

Two remarks before to conclude:

The ASM rebalance activity is not recorded into the v$asm_disk_iostat view. It is recorded into the v$asm_disk_stat view. So, if you are using the asm_metrics utility, you have to change the asm_feature_version variable to a value > your ASM instance version.
I tested with compatible.asm set to 10.1 and 11.2.0.2 and observed the same behaviour for all those metrics.

Conclusion of Part II:

Nothing surprising except (at least for me) that the Avg By/Read is stucked to 1MB (While the allocation unit is 4MB).
I’ll update this post with ASM 12c results as soon as I can (if something new needs to be told).