topic error: proc: table is full in Operating System - HP-UX

error: proc: table is full

Jairo Campana — Mon, 18 Oct 2004 15:30:47 GMT

Hello , I obtain a messages in syslog and dmesg

# dmesg
Oct 18 17:27
...
: table is full
proc: table is full
proc: table is full

more syslog.log
Oct 18 17:15:42 cp1 vmunix: proc: table is full
Oct 18 17:28:05 cp1 above message repeats 4 times

Re: error: proc: table is full

Sridhar Bhaskarla — Mon, 18 Oct 2004 15:34:41 GMT

Hi Jairo,

Your 'nproc' kernel parameter is full. Run 'sar -v 2 20' and observe 'proc-sz' column.

Increase it by around 50%. Also increase 'maxuprc'. Make it a formula like 'nproc -2'.

'maxuprc' overflows errors are not shows in dmesg or syslog. Processes fail with the error EAGAIN.

-Sri

Re: error: proc: table is full

doug mielke — Mon, 18 Oct 2004 15:44:14 GMT

yes, you may need to increase nproc if there is a legitamte need for the processes. If they are the reult of some runaway spawning process ( a programmer test, for instance), but in the meantime, find some processes to kill.
Idle shells, stop and start old non critical apps, etc.

Re: error: proc: table is full

Jairo Campana — Mon, 18 Oct 2004 15:58:35 GMT

Parameter setting
nproc 276 276 Static N/A M

sar -v 2 20

cp1# sar -v 2 20

HP-UX cp1 B.11.00 U 9000/800 10/18/04

17:53:12 text-sz ov proc-sz ov inod-sz ov file-sz ov
17:53:14 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:16 N/A N/A 251/276 0 476/476 0 1102/2510 0
17:53:18 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:20 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:22 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:24 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:26 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:28 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:30 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:32 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:34 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:36 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:38 N/A N/A 251/276 0 476/476 0 1103/2510 0
17:53:40 N/A N/A 251/276 0 476/476 0 1105/2510 0
17:53:42 N/A N/A 260/276 2 476/476 0 1122/2510 0
17:53:44 N/A N/A 260/276 4 476/476 0 1122/2510 0
17:53:46 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:48 N/A N/A 251/276 0 476/476 0 1104/2510 0
17:53:50 N/A N/A 252/276 0 476/476 0 1104/2510 0
17:53:52 N/A N/A 251/276 0 476/476 0 1105/2510 0

maxuprc 75 75 Static N/A

... change the parmeters nproc and maxuprc for new value

Re: error: proc: table is full

Sridhar Bhaskarla — Mon, 18 Oct 2004 16:02:02 GMT

Hi Jairo,

251/276 - obviously it is running low and at anytime you will (you already had) run out of this limit. 251 is the usage and 276 is the configured maximum. I would go ahead and double the value. Keep maxuprc at 'nproc -2' formula as you don't have to keep on changing it anytime you change nproc parameter. Since nproc may affect the number of files opened, you may also want to configure your 'nfile' to a bit higher like 3096.

-Sri

Re: error: proc: table is full

Uday_S_Ankolekar — Mon, 18 Oct 2004 16:06:00 GMT

You got to increasse value of nproc and also the value of maxusers. When you change these value new kernel will be rebuilt and system will reboots

-USA

Re: error: proc: table is full

Michael Tully — Mon, 18 Oct 2004 16:40:21 GMT

The nproc parameter is extremely low. I'd suggest you increase it to at least 2048, also take into account maxuprc and maxusers. You could use formulas as suggested by Sri.

Re: error: proc: table is full

Jairo Campana — Mon, 18 Oct 2004 18:17:11 GMT

thanks for all masters .

my problem is solved

Re: error: proc: table is full

Jairo Campana — Mon, 18 Oct 2004 18:27:17 GMT

changes:

* Tunable parameters

STRMSGSZ 65535
maxswapchunks 512
maxuprc 150
maxusers 150
nfile 2500
nstrpty 60
num_tachyon_adapters 5

Re: error: proc: table is full

Sridhar Bhaskarla — Mon, 18 Oct 2004 18:29:32 GMT

Well - I am not seeing any changes. You should have 'nproc' listed in there. Set your maxuprc to 'nproc -2'. Your 'nfile' is still at 2500??.

-Sri

Re: error: proc: table is full

Robert L. Parris — Tue, 02 Aug 2005 15:53:29 GMT

Are there any warning messages prior to getting the above message? I am working to have OVO monitor this particular issue, but need to notify prior to this event happening, as well as when it happens.