cancel
Showing results for 
Search instead for 
Did you mean: 

Ingress running like a dog!

SOLVED
Go to solution

Ingress running like a dog!

Hi Everyone,

I wonder if anyone can help.

I recently purchased an rp2470 1 x 650MHz CPU 1GB Mem (est 16k tpm). I am using this as a development server for Ingress however it is running like a dog!

I think that I have all the correct OS and application patches!

To make matters worse I installed the same software on an rp5430 2 x 550Mhz CPU (est 18k tpm) and it works about 6 times faster!!!

Does anyone know what is going on? I think that it might be because it is on a 1 CPU server... or memory??

When doing a query the CPU is maxing out!

Thanks in advance!
If you don't ask you don't get!
11 REPLIES
Chris Wilshaw
Honored Contributor

Re: Ingress running like a dog!

We have similar issues running Ingres II under HP-UX 11i, on an rp7410 server.

We currently have a call open with CA regarding this issue (dbms starts to run at 100% of one CPU, and effectively locks out all transactions until the database processes are re-started) - we have been waiting for a patch for a few of months now.

Re: Ingress running like a dog!

Do you have any more information? I would really appreciate it if you could let me know what happens.

Cheers
If you don't ask you don't get!
Donald Kok
Respected Contributor

Re: Ingress running like a dog!

Hi,

I have 2470's with 1Gb on 11.0 and 11i. Ingres works like a charm.

11.0: ingres version II 2.5/0011 (hpb.us5/00)
11.i: ingres version II 2.0/0001 (hpb.us5/00)
(which are not current verions, but why change a winning team)

Good luck
Donald
My systems are 100% Murphy Compliant. Guaranteed!!!
Mark Treen_1
Advisor

Re: Ingress running like a dog!

Maybe not quite what you need but some nice tuning things in here!!
Hope it's of some use?

http://support.ca.com/techbases/ingres/ingres_manuals/ingres_docs.html

Mark
Mark Treen
Chris Wilshaw
Honored Contributor
Solution

Re: Ingress running like a dog!

From the details on our call with CA;

Bug 107914 talks about CPU hangs with internal threads, cache sharing and group buffers.

This is supposedly fixed in newer patches, although we have recently applied patch 9154, and have still seen the problems.

We have been running Ingres II under HP-UX 11.0 on a variety of R, D and L class servers, and have not encountered this problem - it appears to be related to either HP-UX 11i, or the higher spec machine.
Chris Wilshaw
Honored Contributor

Re: Ingress running like a dog!

We have now had an update from CA on our bug.

It is believed that this is fixed in patch9374 (from "Ingres II Corrections" portion of the related doc file)

(GENERIC)
In a multiple dbms installation with internal thread type and group buffers, running large update transactions, a dbms server may appear to hang with 99% cpu. (107914)

As we have only received the patch today, I am unable to confirm at this time if the problem is definitely fixed, but I will pass this information on as soon as possible.

Chris
Mike Laytham_1
Occasional Visitor

Re: Ingress running like a dog!


Has the patch supplied by CA fixed the problem. We have a similar scenario, where we have multiple DBMS servers and when running large updates the DBMS servers all go to the top of the CPU utilization process list.

The servers don't hang but simply eat up CPU.

Thanks

Mike
Chris Wilshaw
Honored Contributor

Re: Ingress running like a dog!

That particular patch didn't work, but we have had another update (patch 9374).

This has been on test for a week or so, with no sign of problems.

From the patch doc

(GENERIC)
In a multiple dbms installation with internal thread type
and group buffers, running large update transactions,
a dbms server may appear to hang with 99% cpu.
(107914)

It also fixes another "known" issue that CA informed us of which they refer to as the "query-flattening" problem. This was an issue that we encountered a couple of times, and that we were able to reproduce at will (rather than the server hanging problem which appeared to be random).
Mike Laytham_1
Occasional Visitor

Re: Ingress running like a dog!

Thanks Chris.

Can you give an idea of how your problem manifests itself. We have Ingres installed on a muli processor machine running 7 DBMS servers with over 3,600 connections. When we have peak load and a large amount of updates taking place with Ingres, the DBMS servers all go into a 'high cpu' state and sit at the top of the process table. We can still interact with the DBMS servers 'albeit extremely slowly' and if we remove around 150 sessions they seem to recover.
However we are running OS threads on a TRU64 machine, so Im trying to ascertain if the symptoms appear to be similar, and are related to Ingres, rather than the flavour of Unix.

Thanks

Mike
Mike Laytham_1
Occasional Visitor

Re: Ingress running like a dog!

Has the patch fixed the problem, as we are still in dispute with CA as to who's issue our problem resides with... OS or Ingres..

If you could supply some more information on what happens it would be appreciated.

thanks

Mike
Chris Wilshaw
Honored Contributor

Re: Ingress running like a dog!

Mike,

So far, we've had no problems on the patched systems.

The issues that we had were;

1) "spinning" DBMS server. Some unknown process attaches to the dbms server, causing it to take up as much CPU as possible. Once this happens, no other processes are able to connect to the server (sql, ipm etc all just hang). The only solution here is to kill off all application processes, then try to shut ingres down. Most of the time this works, although occasionally, you have to kill off a couple of the ingres processes too.

2) query flattening problem. This is supposedly due to an SQL routine picking up an inefficient query plan when it runs. When this happens, other processes can connect to the server, but everything takes a long time to run. Once the initial query ends (or is killed), the server state returns to normal. This sounds like it fits your problem.

The spinning server was our main problem (we were getting one of those a week on average). We only had 2 instances of the query flattening problem, and they were fixed by the application developers by setting (I think) trace points within the code to force the choice of query plan.