The "steal" you are showing here is at the LPAR level where workload is
running on an LPAR logical engine, but that engine is also being
utilized by the other LPAR, so one of them must wait (measured as
"steal").  This is to be expected at high utilization.  I prefer to look
at the summary that looks something like the following, showing your 10
engines and how they are utilized.  The two Linux/IFL LPARs are running
for a total above 90% of available resource.  When running above 90%,
there is just going to be a lot of CPU waits (shown on ESAXACT as Bill
said), and "steals" as measured by linux and the LPAR.  This is only a
problem if your backups are not completing on time, otherwise, it is
just telling you that work would complete faster if there was more
resource available (more IFLs).  Alternatively, reduce the weight for
VML4, and that will give more resource to VML2.

Totals by Processor type:
<---------CPU-------> <-Shared Processor busy->
Type Count Ded shared  Total  Logical Ovhd Mgmt
---- ----- --- ------ ------ -------- ---- ----
CP       2   0      2  193.3    184.5  2.4  6.4
IFL      4   0      4  394.3    393.2  0.4  0.7
ICF      3   3      0    0.5        0    0  0.5
ZIIP     1   0      1    8.5      7.9  0.1  0.5






On 2/1/2018 11:58 AM, Victor Echavarry wrote:
Barton:

This is the report of  the CPU at the moment of the steal, VML2 is production 
and VML4 is test.

ESALPAR - Logical Partition Analysis - VML2
          <--Complex--> <---Logical---->              <-- Logical Processor --> 
<--------CPU (percentages)-------> <Multi-thread>
          Phys Dispatch <--Partition---> <-CPU--> CPU <%Assigned>       Cap  Wt 
Total  Emul  User   Sys  Idle  Stl  Idle  cp1/cp2
Time     CPUs    Slice Nr Name     Type Cnt Type  ID  Total Ovhd  Wght ped Cmp  
util  time ovrhd ovrhd  time  Pct  Time
-------- ---- -------- -- -------- ---- --- ---- --- ------ ---- ----- --- --- 
----- ----- ----- ----- ----- ---- ------ --- ---
20:21:00   10  Dynamic 01 VML2     IFL    4 CPU  Tot  253.3  0.2    75  No  No 
252.7 238.5   8.8   5.3        147
20:21:00   10  Dynamic 03 VML4     IFL    3 CPU  Tot  121.6  1.2    60 Yes  No
20:20:00   10  Dynamic 01 VML2     IFL    4 CPU  Tot  255.0  0.2    75  No  No 
254.5 240.7   8.4   5.4        146
20:20:00   10  Dynamic 03 VML4     IFL    3 CPU  Tot  119.7  1.1    60 Yes  No
20:19:00   10  Dynamic 01 VML2     IFL    4 CPU  Tot  250.3  0.3    75  No  No 
249.8 235.9   8.5   5.4        150
20:19:00   10  Dynamic 03 VML4     IFL    3 CPU  Tot  125.2  1.6    60 Yes  No
20:18:00   10  Dynamic 01 VML2     IFL    4 CPU  Tot  257.1  0.2    75  No  No 
256.5 242.2   8.8   5.6        144
20:18:00   10  Dynamic 03 VML4     IFL    3 CPU  Tot  114.6  1.2    60 Yes  No

The majority of the steal issues occurs at night. At that time backups process 
starts.

Regards,

Victor Echavarry
System Programmer
Operating Systems
EVERTEC, LLC



-----Original Message-----
From: Linux on 390 Port [mailto:LINUX-390@VM.MARIST.EDU] On Behalf Of barton
Sent: Thursday, February 1, 2018 3:24 PM
To: LINUX-390@VM.MARIST.EDU
Subject: Re: Guest show steal and the LPAR has available CPU

It's very important to look at it first from the total IFL perspective.
What is the total IFL utilization?  Then break that down by LPAR.
If 100% of your IFLs are in use, and one LPAR is using 60% of the IFL resource 
and the other LPAR is using 40% of the IFL resource, then it's a matter of 
prioritizing the workload.  Relative share of 100 means there is no 
prioritization assigned.
If the guests need more CPU, then they need a larger share.  If the LPAR needs 
more CPU, it needs a larger weight...

On 2/1/2018 9:47 AM, Victor Echavarry wrote:
We have a situation that a couple of guest has CPU stealing and when we 
checking the LPAR, the IFL have available CPU for processing. What could be 
possible? The guest sharing are based on CPU assigned, for example a guest with 
1 CPU has 100 relative share and a 2 CPU has 200 relative share. We are running 
under z/VM 6.4 and the SLES is 11SP4.

Regards,

Victor Echavarry

System Programmer

Operating Systems

EVERTEC, LLC








----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions, send
email to lists...@vm.marist.edu with the message: INFO LINUX-390 or
visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/


----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions, send email to 
lists...@vm.marist.edu with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit http://wiki.linuxvm.org/

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/



----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to