oninit use 95+%cpu
Posted in 2000
Topics: Platform-Specific Issues, Versions, Editions & End-of-Life
Hi,
We have a production server which is running IDS 7.31uc2 on HP-UX 10.20. This
morning one oninit process use 95+% cpu and none of the users could do anything.
No any error could find in informix online log and unix syslog.
At that time, there was a informix trainer happened to be on site and he suggest
changed two parameter:
1)SERVERNUM from "101" to be "1". His explanation is that because when it was
101, the shared memory was loosing contiguous space. He says that when the
default is 1 then contiguous space it retained.
2) RESIDENT from NO(0) to YES(1).
I am not sure about these two parameters. As my understanding, the value is not
important. It just need to be unique to the instance. And this machine, totally
256MB physical memory and informix use 110MB. For me it is too tight to
resident.
Anybody can tell me if these two parameters related to the problem or explain to
me the reason?
Thanks in advance.
Vivian
vyao@dehavilland.ca
vyao@dehavilland.ca wrote:
>
> Hi,
>
> We have a production server which is running IDS 7.31uc2 on HP-UX 10.20. This
> morning one oninit process use 95+% cpu and none of the users could do anything.
> No any error could find in informix online log and unix syslog.
>
> At that time, there was a informix trainer happened to be on site and he suggest
> changed two parameter:
> 1)SERVERNUM from "101" to be "1". His explanation is that because when it was
> 101, the shared memory was loosing contiguous space. He says that when the
> default is 1 then contiguous space it retained.
Umm, no. Just no. SERVERNUM has EXACTLY one effect it is added to the base
Informix shared memory key to create a key unique to that particular
instance. Believe me when I say that NO UNIX will allocate shared memory
differently if the key is 0x52534801 rather than 0x52b74801!
> 2) RESIDENT from NO(0) to YES(1).
This WILL effect performance but I do not think it would cause the oninit to
hog the CPU, on the contrary I think it would cause the oninits to be
suspended while memory was swapped back in if any problem were to result.
> I am not sure about these two parameters. As my understanding, the value is not
> important. It just need to be unique to the instance. And this machine, totally
> 256MB physical memory and informix use 110MB. For me it is too tight to
> resident.
A little tight but not overly so. Depends on what else is running on that
machine.
> Anybody can tell me if these two parameters related to the problem or explain to
> me the reason?
I can. They do not. What does is another story. Try posting some onstat
output and your ONCONFIG file. Recommended reports: -g ath, -g glo, -g seg,
-p, -d, -l, and about 20 lines of the message log from around the worst time.
Art S. Kagel
I've seen this once, in my case it was caused by someone writing some SQL
joining two tables, about 4000 and 100000 rows, and forgetting to put the
join in the where clause. This put the engine in some sort of super tight
loop, I could not even get statistics out of it to see what was happening,
When I saw his screen and quit the SQL all returned to normal and I got to
shout for a while :-/
--
---------------------------------------
Tony Flaherty aef@mfs.misys.co.uk
Analyst Programmer
Misys Financial Systems
All statements and opinions are my own,
Misys don't pay me enough to have opinions
on their behalf
.
vyao@dehavilland.ca wrote in message <86ij0t$4e2$1@news.xmission.com>...
>
>
>
>Hi,
>
>We have a production server which is running IDS 7.31uc2 on HP-UX 10.20.
This
>morning one oninit process use 95+% cpu and none of the users could do
anything.
>No any error could find in informix online log and unix syslog.
>
>At that time, there was a informix trainer happened to be on site and he
suggest
>changed two parameter:
>1)SERVERNUM from "101" to be "1". His explanation is that because when it
was
>101, the shared memory was loosing contiguous space. He says that when
the
>default is 1 then contiguous space it retained.
>2) RESIDENT from NO(0) to YES(1).
>
>I am not sure about these two parameters. As my understanding, the value is
not
>important. It just need to be unique to the instance. And this machine,
totally
>256MB physical memory and informix use 110MB. For me it is too tight to
>resident.
>
>Anybody can tell me if these two parameters related to the problem or
explain to
>me the reason?
>
>Thanks in advance.
>
>Vivian
>vyao@dehavilland.ca
>
>