/scripts/ahub.form.attachments.js
0

TF 753 or 757 ASE 16

Nov 24, 2016 at 04:03 PM

214

avatar image

Hello,

ASE 16 SP02 PL03 on RHEL 7.2

In order to reduce or get rid of the message "Spinlock Resource->rproccache_spin at address...", "Procedure Cache Spinlock (Resource->rproccache_spin)", ect

which TF should be used : TF 753 or TF 757 ?

I was checking https://wiki.scn.sap.com/wiki/display/SYBASE/Spinlocks+and+CPU+usage+in+SAP+ASE#

and I was just thinking which one it would be best.. or both ?

Thank you

Jose

10 |10000 characters needed characters left characters exceeded
* Please Login or Register to Answer, Follow or Comment.

3 Answers

Best Answer
Mark A Parsons Nov 24, 2016 at 04:57 PM
0

See the following for discussions re: procedure cache (spinlocks), ELC and TFs 753/757:

ASE Procedure Cache and Engine Local Cache (ELC)

Get Rid of the Crutches – Right Size Proc Cache

Share
10 |10000 characters needed characters left characters exceeded
Mike Willett Nov 24, 2016 at 05:59 PM
0

I think TF753 has been superseded by the config setting

"enable large chunk elc" in Sybase 16

We have it set to 0 - although I'm told Sybase 16 works better than Sybase 15.7 with large chunk elc.

However, we haven't tried it so can't comment how well it works.

Show 1 Share
10 |10000 characters needed characters left characters exceeded

Hi Mike,

In 16.0:

TF 753 is replaced by "large allocation auto tune"
TF 758 is replaced by "enable large chunk elc"

0
Jose Torres Nov 24, 2016 at 07:06 PM
0

Thank you guys,

But still wondering:

-Is TF 757 is similar to TF753 ?[( if any one has the specifics on 757 it'll be great)

-According to "ASE Procedure Cache and Engine Local Cache (ELC)"

large allocation auto tune (replaces traceflag 753) -> we have set it to 0 for a while, but we are still getting this message followed by a stack trace:

"Spinlock Resource->rproccache_spin at address..."

then ,It's advisable to use 753 and eventually increase Proc cache ?

-Does TF 757 has been replaced by a new ASE parameter ?

Regards

Jose

Show 2 Share
10 |10000 characters needed characters left characters exceeded

Correction on my response : We have set "enable large chunk elc" to 0. So "large allocation auto tune" would it be advisable to switch it to 1?

Thank you

Jose

0

Hi Jose,

Keep in mind, the association with any of these trace flags and configurations have no bearing on whether or not normal processing requires rproccache_spin spinlock. These configurations change usage of the caches not the necessity of the spinlock when it is being used.

Asking how to get rid of "Spinlock Resource" issues during a stacktrace, is answered by getting rid of the stack trace. Spinlocks will be held anytime a protected memory address needs to be accessed. As long as there are multiple engines, multiple users competing for the same memory, you will have grabs as well as spins/waits for the spinlock.

1
Skip to Content