I have two LPARS, both running Oracle for the same application (one is
Test, the other is Development). Each has two processors assigned,
and because they're dual-core, they show up in NMON as four
processors. On one LPAR, it uses all four processors equally. On the
other, it appears to only be using two of the processors (the other
two show 100% idle all the time). I've never seen that behavior
before, either from an LPAR of a physical box. What might cause that?

rodak schrieb:
> I have two LPARS, both running Oracle for the same application (one is
> Test, the other is Development). Each has two processors assigned,
> and because they're dual-core, they show up in NMON as four
> processors. On one LPAR, it uses all four processors equally. On the
> other, it appears to only be using two of the processors (the other
> two show 100% idle all the time). I've never seen that behavior
> before, either from an LPAR of a physical box. What might cause that?
Dual-core will not make 4 out of 2.... this is SMT.
It seems on that 2nd LPAR you don't have enough load
to use that 2nd thread on the two CPUs, so they stay
idle.

There would seem to be plenty of load on the system - processors 0 and
1 are running 80-100% busy, while processors 2 and 3 are dead-idle.
What would cause that?

On Aug 28, 1:15*am, Thomas Braunbeck
wrote:
> rodak schrieb:> I have two LPARS, both running Oracle for the same application (one is
> > Test, the other is Development). *Each has two processors assigned,
> > and because they're dual-core, they show up in NMON as four
> > processors. *On one LPAR, it uses all four processors equally. *On the
> > other, it appears to only be using two of the processors (the other
> > two show 100% idle all the time). *I've never seen that behavior
> > before, either from an LPAR of a physical box. *What might cause that?
>
> Dual-core will not make 4 out of 2.... this is SMT.
> It seems on that 2nd LPAR you don't have enough load
> to use that 2nd thread on the two CPUs, so they stay
> idle.

>
> output of "lparstat -i" on the two lpars in question would be helpful.
First is the one that uses all four processors more or less equally:

Node Name : HQDBLAW2
Partition Name : HQDBLAW2
Partition Number : 5
Type : Shared-SMT
Mode : Capped
Entitled Capacity : 2.00
Partition Group-ID : 32773
Shared Pool ID : 0
Online Virtual CPUs : 2
Maximum Virtual CPUs : 2
Minimum Virtual CPUs : 2
Online Memory : 4096 MB
Maximum Memory : 8192 MB
Minimum Memory : 2048 MB
Variable Capacity Weight : 0
Minimum Capacity : 2.00
Maximum Capacity : 2.00
Capacity Increment : 0.01
Maximum Physical CPUs in system : 32
Active Physical CPUs in system : 16
Active CPUs in Pool : 16
Unallocated Capacity : 0.00
Physical CPU Percentage : 100.00%
Unallocated Weight : 0

Sample NMON output:
nmon 2=Top-Child-CPU Host=HQDBLAW2 Refresh=2 secs
14:22.37
CPU-Utilisation-Small-View EntitledCPU= 2.00 UsedCPU=
1.162
Logical CPUs
0----------25-----------50----------75----------100
CPU User% Sys% Wait% Idle%| | |
| |
0 19.9 13.9 2.0 64.2|UUUUUUUUUssssss
> |
1 13.9 20.4 0.5 65.2|UUUUUUssssssssss
> |
2 10.0 23.9 1.5 64.7|UUUUsssssssssss
> |
3 25.9 30.8 1.0 42.3|UUUUUUUUUUUUsssssssssssssss
> |
EntitleCapacity/VirtualCPU
+-----------|------------|-----------|------------+
EC 24.4 23.1 0.2 10.5|
UUUUUUUUUUUUsssssssssssiiiii---------------------|
VP 24.4 23.1 0.2 10.5|
UUUUUUUUUUUUsssssssssssiiiii---------------------|
EC= 58.1% VP= 58.1% +--Capped---|------------|-----------100%
VP=2 CPU+


Next is the one that only seems to want to use two of the processors:

Node Name : HQDBLAW3
Partition Name : HQDBLAW3
Partition Number : 3
Type : Shared-SMT
Mode : Capped
Entitled Capacity : 2.00
Partition Group-ID : 32771
Shared Pool ID : 0
Online Virtual CPUs : 2
Maximum Virtual CPUs : 2
Minimum Virtual CPUs : 2
Online Memory : 4096 MB
Maximum Memory : 8192 MB
Minimum Memory : 2048 MB
Variable Capacity Weight : 0
Minimum Capacity : 2.00
Maximum Capacity : 2.00
Capacity Increment : 0.01
Maximum Physical CPUs in system : 32
Active Physical CPUs in system : 16
Active CPUs in Pool : 16
Unallocated Capacity : 0.00
Physical CPU Percentage : 100.00%
Unallocated Weight : 0

sample NMON output, while three DB report jobs are running:
CPU-Utilisation-Small-View EntitledCPU= 2.00 UsedCPU=
0.649
Logical CPUs
0----------25-----------50----------75----------100
CPU User% Sys% Wait% Idle%| | |
| |
0 34.0 19.0 9.5 37.5|
UUUUUUUUUUUUUUUUUsssssssssWWWW> |
1 18.5 8.0 4.0 69.5|UUUUUUUUUssssWW
> |
2 0.0 0.0 0.0 100.0|
> |
3 0.0 0.0 0.0 100.0|
> |
EntitleCapacity/VirtualCPU
+-----------|------------|-----------|------------+
EC 21.4 7.2 0.4 3.4|
UUUUUUUUUUsssi-----------------------------------|
VP 21.4 7.2 0.4 3.4|
UUUUUUUUUUsssi-----------------------------------|
EC= 32.4% VP= 32.4% +--Capped---|------------|-----------100%
VP=2 CPU+

rodak wrote:
>> output of "lparstat -i" on the two lpars in question would be helpful.
> First is the one that uses all four processors more or less equally:
>
> Node Name : HQDBLAW2
> Partition Name : HQDBLAW2
> Partition Number : 5
> Type : Shared-SMT
> Mode : Capped
> Entitled Capacity : 2.00
> Partition Group-ID : 32773
> Shared Pool ID : 0
> Online Virtual CPUs : 2
> Maximum Virtual CPUs : 2
> Minimum Virtual CPUs : 2
> Online Memory : 4096 MB
> Maximum Memory : 8192 MB
> Minimum Memory : 2048 MB
> Variable Capacity Weight : 0
> Minimum Capacity : 2.00
> Maximum Capacity : 2.00
> Capacity Increment : 0.01
> Maximum Physical CPUs in system : 32
> Active Physical CPUs in system : 16
> Active CPUs in Pool : 16
> Unallocated Capacity : 0.00
> Physical CPU Percentage : 100.00%
> Unallocated Weight : 0
>
> Sample NMON output:
> nmon 2=Top-Child-CPU Host=HQDBLAW2 Refresh=2 secs
> 14:22.37
> CPU-Utilisation-Small-View EntitledCPU= 2.00 UsedCPU=
> 1.162
> Logical CPUs
> 0----------25-----------50----------75----------100
> CPU User% Sys% Wait% Idle%| | |
> | |
> 0 19.9 13.9 2.0 64.2|UUUUUUUUUssssss
>> |
> 1 13.9 20.4 0.5 65.2|UUUUUUssssssssss
>> |
> 2 10.0 23.9 1.5 64.7|UUUUsssssssssss
>> |
> 3 25.9 30.8 1.0 42.3|UUUUUUUUUUUUsssssssssssssss
>> |
> EntitleCapacity/VirtualCPU
> +-----------|------------|-----------|------------+
> EC 24.4 23.1 0.2 10.5|
> UUUUUUUUUUUUsssssssssssiiiii---------------------|
> VP 24.4 23.1 0.2 10.5|
> UUUUUUUUUUUUsssssssssssiiiii---------------------|
> EC= 58.1% VP= 58.1% +--Capped---|------------|-----------100%
> VP=2 CPU+
>
>
> Next is the one that only seems to want to use two of the processors:
>
> Node Name : HQDBLAW3
> Partition Name : HQDBLAW3
> Partition Number : 3
> Type : Shared-SMT
> Mode : Capped
> Entitled Capacity : 2.00
> Partition Group-ID : 32771
> Shared Pool ID : 0
> Online Virtual CPUs : 2
> Maximum Virtual CPUs : 2
> Minimum Virtual CPUs : 2
> Online Memory : 4096 MB
> Maximum Memory : 8192 MB
> Minimum Memory : 2048 MB
> Variable Capacity Weight : 0
> Minimum Capacity : 2.00
> Maximum Capacity : 2.00
> Capacity Increment : 0.01
> Maximum Physical CPUs in system : 32
> Active Physical CPUs in system : 16
> Active CPUs in Pool : 16
> Unallocated Capacity : 0.00
> Physical CPU Percentage : 100.00%
> Unallocated Weight : 0
>
> sample NMON output, while three DB report jobs are running:
> CPU-Utilisation-Small-View EntitledCPU= 2.00 UsedCPU=
> 0.649
> Logical CPUs
> 0----------25-----------50----------75----------100
> CPU User% Sys% Wait% Idle%| | |
> | |
> 0 34.0 19.0 9.5 37.5|
> UUUUUUUUUUUUUUUUUsssssssssWWWW> |
> 1 18.5 8.0 4.0 69.5|UUUUUUUUUssssWW
>> |
> 2 0.0 0.0 0.0 100.0|
>> |
> 3 0.0 0.0 0.0 100.0|
>> |
> EntitleCapacity/VirtualCPU
> +-----------|------------|-----------|------------+
> EC 21.4 7.2 0.4 3.4|
> UUUUUUUUUUsssi-----------------------------------|
> VP 21.4 7.2 0.4 3.4|
> UUUUUUUUUUsssi-----------------------------------|
> EC= 32.4% VP= 32.4% +--Capped---|------------|-----------100%
> VP=2 CPU+





looks OK to me so far. Does "smtctl" output on the 2 lpars differ?
Has oracle been configured identically? (not my field)

Regards
Joachim Gann

Yes, smtctl produces the same output on both:
-------------------------------------------------------------------------------
This system is SMT capable.
SMT is currently enabled.
SMT boot mode is not set.
SMT threads are bound to the same virtual processor.

proc0 has 2 SMT threads.
Bind processor 0 is bound with proc0
Bind processor 1 is bound with proc0

proc2 has 2 SMT threads.
Bind processor 2 is bound with proc2
Bind processor 3 is bound with proc2
-------------------------------------------------------------------------------

I presume Oracle is configured identically - my DBA is pretty
persnickety about such things.

But maybe not - I just fired of a bunch of simple "find" commands, and
saw from nmon that it did use all four processors. I've asked my DBA
if there's something in Oracle that determines how it handlse multi-
threaded processes.

On Aug 28, 2:50*pm, Joachim Gann wrote:>
> looks OK to me so far. Does "smtctl" output on the 2 lpars differ?
> Has oracle been configured identically? (not my field)
>
> Regards
> Joachim Gann

rodak schrieb:
>> output of "lparstat -i" on the two lpars in question would be helpful.
> First is the one that uses all four processors more or less equally:

> CPU-Utilisation-Small-View
> EntitledCPU= 2.00
> UsedCPU= 1.162


>
> Next is the one that only seems to want to use two of the processors:
>

> CPU-Utilisation-Small-View
> EntitledCPU= 2.00
> UsedCPU= 0.649

The 2nd system uses less than 1 CPU, so why should it touch the
2nd CPU at all? Put sole load on that box (just a looping script
will do) and see how the 2nd CPU starts 'working' when the UsedCPU
value gets > 1