TNS-00510: Internal limit restriction exceeded HPUX Error: 24: Too many open files

[likid0]

With sar -v the third column is your nfile use:

#sar -v

00:00:01 text-sz ov proc-sz ov inod-sz ov file-sz ov
00:15:00 N/A N/A 246/4200 0 1899/162600 0 3490/65536 0
00:30:01 N/A N/A 262/4200 0 1911/162600 0 3525/65536 0
00:45:00 N/A N/A 248/4200 0 1899/162600 0 3493/65536 0
01:00:00 N/A N/A 259/4200 0 1912/162600 0 3522/65536 0
01:15:01 N/A N/A 249/4200 0 1902/162600 0 3500/65536 0

[Johnson Punniyalingam]

Hi All,

As discussed earlier have increased "nfile" from "3000" to "60000" but still have below problem,

so can any one advice no i should also tune

"maxfile" & "maxfiles_lim"

Oracle listener fail again at 17:40 pm.
Below our listener log.

27-MAY-2009 17:41:16 * 1159
TNS-01159: Internal connection limit has been reached; listener has shut down
TNS-12540: TNS:internal limit restriction exceeded
TNS-12560: TNS:protocol adapter error
TNS-00510: Internal limit restriction exceeded
HPUX Error: 24: Too many open files
No longer listening on: (DESCRIPTION=(ADDRESS=(PROTOCOL=ipc)(KEY=EIFSDW)))
No longer listening on: (DESCRIPTION=(ADDRESS=(PROTOCOL=tcp)(HOST=dwhsvr01)(PORT=1521)))


Just for your info, I noticed that the ulimit in dwhsvr01 and dwhsvr02 a bit different.


/home/oracle > ulimit -a
time(seconds) unlimited
file(blocks) unlimited
data(kbytes) 262144
stack(kbytes) 8192
memory(kbytes) unlimited
coredump(blocks) 4194303
nofiles(descriptors) 60

[Johnson Punniyalingam]

sorry... "typ0 error" we changed it " 6000 " for nfile . still the problem comes again listerner down.

Thx,
Johnson

[Rita C Workman]

Would you do me a favor, cause I tend to shy away from recommending parm changes without knowing what the box has to offer in resources....so please give me some info:

O/S version?
Oracle version?
Any other appl software running on it?
# of CPU's?
Amt of RAM?
Is this in a clustered environment?
Can other apps/pkgs fail to this box?

You gave us the nfile setting

dbc_max_% ? or higher os filecache_max
dbc_min_% ? or higher OS filecache_min
nproc
ninode
maxdsiz (32 & 64 bit)
maxssiz "
maxtsiz "
vx_ninode
..and all your semaphore parm settings..

Thanks,
Rita

[James R. Ferguson]

Hi Johnson:

While I realize that your release is 11.11, the manpages beginning at 11.23 include excellent documentation on many, many kernel tunables. I suggest that this may be very useful to you.

http://docs.hp.com/en/B3921-60631/index.html

Regards!

...JRF...

[Johnson Punniyalingam]

Hi Rita,

Bunch of thanks for you concern. please see below information as requested.

Parameter Current Dyn Planned Module Version
===============================================================================

dbc_max_pct 50 - 50
dbc_min_pct 5 - 5

nproc 1620 - (20+8*MAXUSERS)
ninode 1988 - ((NPROC+16+MAXUSERS)+32+(2*NPTY))
maxdsiz 0x10000000 - 0x10000000
maxdsiz_64bit 0x40000000 - 0X40000000
vx_ninode 0 - 0
sema 1 - 1
semaem 16384 - 16384
semmap 202 - (SEMMNI+2)
semmni 200 - 200
semmns 400 - 400
semmnu 30 - 30
semmsl 2048 Y 2048
semume 10 - 10
semvmx 32767 - 32767

O/S version?
OS: B.11.11,

Oracle version?
Oracle 8

Any other appl software running on it?

NO

# of CPU's?

CPU: 4 x 550 MHz, PA-RISC2.0, v.


Amt of RAM?

RAM: 12288 MB

Is this in a clustered environment?

NO

Can other apps/pkgs fail to this box?

NO cluster ENV.

[Rita C Workman]

OK...4 CPU, 12Gb RAM, Oracle 8..and not knowing your shop - here's a few suggestions:

dbc_max_% - chg to 10 to 13
dbc_min_% - chg to 5 to 8
I tend to go to the higher number, but you pick. YOur settings are defaults-BAD defaults!!

nproc 1620 - maybe higher say 2000 or even 4000 if you run big batch jobs

ninode 1988 - I'd up to 2048 - 4096. This parm creates a recycle table for inodes. The kernel calc creates way way to big a table that the sar command shows you don't need.

vx_ninode 0 - I'd hard set this to 20m to 40m. Leaving it at zero the system then creates a huge table of wasted resources on memory.You can increase if you need to, but 40m is plenty.

...on your semaphores...probably too small, can use sar -mS 1 10 (or other intervals/counts) to monitor...
sema 1 - 1
semaem 16384 - 16384
semmap 202 - try 2002 or 4002
semmni 200 - try 2000 or 4000
semmns 400 - try 2000 or 4000
semmnu 30 - try 1000 or 1500
semmsl 2048 Y 2048 - ok
semume 10 - 10 - try 1000
semvmx 32767 - 32767 - ok

I didn't address the maxd parm, cause I'd like to know what you have your shmmax set at first. So let us know.
A hint - just because Oracle says you need to set a parm to a certain value, doesn't mean you have to or even that you should.

Hope this helps. Remember, kernel tuning is not a science it's more an art. And each shop needs to tune for it's needs, so everyone tunes to their own music.

Rita

[Rita C Workman]

Johnson,

I apologize for not reading more closely on one posting above - I thought the sar output above was from your system, but it appears it was Daniels.
So run your own sar -v command.

The three columns give you great info for procs/inodes/files as Daniel mentions. Each column breakdown include 'current usage'/'parm setting'.

/rcw

[Rita C Workman]

Gotta head to a 'seminar' or something shortly, so here goes:

shmmax - I tend to go with something Stephen Ciullo said - give it to them. So for your 12Gb RAM, I'd leave around 2Gb for that system reserve and I'd give them the 10Gb left.
shmmax - 0x280000000 (hope I converted that right - you might want to double check it)

Now the maxdsiz value. If I counted the zeros right yours comes out to (dec)268435456. Looks ok to me.

For your nfile....I agree with what has been said, and tune as needed.

Rgrds,
Rita

[Dennis Handly]

>so can anyone advice no I should also tune maxfile & maxfiles_lim
nofiles(descriptors) 60

I mentioned how to increase them and why your problem wasn't nfile, yet. Why is this still 60?