SEMMNS does not always win.
If we follow your RHEL5 example we can see in the equivalent RHEL5 RedHat
note:
https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/5/htm
l/tuning_and_optimizing_red_hat_enterprise_linux_for_oracle_9i_and_10g_datab
ases/sect-oracle_9i_and_10g_tuning_guide-setting_semaphores-the_semmns_param
eter
That
The maximum number of semaphores that can be allocated on a Linux system
will be the lesser of: SEMMNS or (SEMMSL * SEMMNI).
So if you set SEMMNS larger than (SEMMSL * SEMMNI), you will only get
(SEMMSL * SEMMNI).
No one said to not check out whatever Oracle Support asks you to do.
However, If they ask you to do something that will cause no harm to your
system to solve a problem even though in your own mind the numbers don't
look as right as you may want, you can still perform the task.
1. It could end up not solving the problem and you move on to other possible
solutions.
2. You can always ask support for an explanation if it doesn't make sense to
you. If you get someone good in Support on the Linux side you may learn some
new things. The Australian Linux team is good, so you may look at when you
escalate your SR and call in as to which team you get. Also if you can get
the World Wide Lead, normally based in US CA time zone, except for
Infiniband switches (which is India), they are excellent and really want to
help solve the problems.
3. It could solve the problem. Then you would ask Oracle Support is this the
new best value for the parameters or when will there be a patch to solve the
problem so you are not deviating from default values in the Oracle setup
across your fleet.
As shown in values coming from Oracle Linux off the Exadatas there are
higher values like the ones recommended by Oracle Support to solve this
problem.
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx> On
Behalf Of Clay Jackson (cjackson)
Sent: Saturday, February 13, 2021 4:15 PM
To: dimensional.dba@xxxxxxxxxxx; gogala.mladen@xxxxxxxxx;
oracle-l@xxxxxxxxxxxxx
Subject: RE: linux kernel parameters
OK - I feel compelled to jump in here - I'm nowhere near the DBA Mladen is;
but, to paraphrase, I know a thing or two about Linux because I've seen a
thing or two, INCLUDING what happens if SMMNS is less than SEMMSL*SEMMNI.
Bottom line (at least in Fedora and RHEL 5, which is where I saw this - I'll
admit I haven't "tested" it since"; but, once bitten, twice shy) SEMMNS
wins, always.
See also https://access.redhat.com/solutions/1437013
If SEMMSL*SEMMNI is less than SEMMNS, you'll only get SMMNS semaphores, NOT
SMMSL@SEMMNI. Linux is quite "forgiving" that way, if you really want to
shoot yourself in the foot, it won't stop you.
And trusting w/o verifying what ANY support person tells you is a fool's
errand.
Clay Jackson
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx> On
Behalf Of dimensional.dba@xxxxxxxxxxx
Sent: Saturday, February 13, 2021 1:30 PM
To: gogala.mladen@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: RE: linux kernel parameters
CAUTION: This email originated from outside of the organization. Do not
follow guidance, click links, or open attachments unless you recognize the
sender and know the content is safe.
You are very funny Mladen. I have never said in any post here, that I don't
like you.
I just believe in facts.
The facts do not support your argument and the RedHat/Oracle documentation
doesn't either. The Oracle values that Oracle itself sets on Exadata, a
fully run and managed Oracle device, doesn't support your argument.
Oracle Support has given a recommendation rather you like it or not.
The originally poster can decide to use those values or not to try and
resolve their issue.
There is nothing wrong with those values as Linux will accept them (will not
damage the system), then follow the rules embedded in the source code as to
what it determines or needs for the resource requirements on the system.
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx <oracle-l-bounce@xxxxxxxxxxxxx> On
Behalf Of Mladen Gogala
Sent: Saturday, February 13, 2021 1:11 PM
To: dimensional.dba@xxxxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: Re: linux kernel parameters
Which ones? I have explained the semantics of the semaphore parameters.
Nobody questioned that semantics. The rest of my statements logically
follows that syntax. The statement that SEMMNS should be larger or equal,
idally equal, to SEMMSL*SEMMNI follows from that syntax. If SEMMNS <
SEMMSL*SEMMNI then you will not be able to allosace SEMMNI semaphore sets
with the length of SEMMSL. That is the reason for my objection to 3000 32000
3000 128 advice. That advice was obviously given by a junior support analyst
who doesn't understand semaphores very well because that recommendation
doesn't conform to the semantics of the semaphore parameters.
This feels like whipping a dead horse. We're getting nowhere. I understand
that you don't like me, but the facts are the facts.
On 2/13/21 3:13 PM, dimensional.dba@xxxxxxxxxxx wrote:
It is not about being right, it is about being technically accurate.
You make statements with no facts.
