HPUX 10.20 kernel parms: MSGSSZ & SEMMNU

[John Knapp]

We are running Lawson Financial applications on top of an Oracle v8.05.1 database on a K570 with 3.5 GBs of RAM. Recently we've experienced Lawson application outages with errors "oradb8 RcvSrvMsg: can't receive message Id 277 Interrupted system call"

Lawson support center recommended we increase our MSGSSZ & SEMMNU kernel parameter values from "8" to "128" and from "515" to "1024" respectively. We just did that this afternoon.

My question is ... how can we proactively measure the application use of those kernel parameters to determine if the new values are adequate?

[Pauline Brasch]

Hi John
Do a 'man' on ipcs and see if this fits the bill.
HTH

[John Knapp]

Pauline ... thanks for the tip but I am not absolutely clear on what to look for in the ipcs output .... here is an example:

# ipcs -bcopt | more
IPC status from /dev/kmem as of Mon Feb 4 16:38:37 2002
T ID KEY MODE OWNER GROUP CREATOR CGROUP CBYTES QNUM QBYTES LSPID LRPID STIME RTIME CTIME
Message Queues:
q 485 0x70700000 -Rrw-rw-rw- root lawson root lawson 0 0 7200 5337 2821 16:38:34 16:38:34 14:31:35
q 486 0x70700001 --rw-rw-rw- root lawson root lawson 0 0 72 0 0 no-entry no-entry 14:31:35
q 487 0x70700002 --rw-rw-rw- root lawson root lawson 0 0 36 2842 2841 16:37:01 16:37:01 14:31:35
q 488 0x70700003 --rw-rw-rw- root lawson root lawson 0 0 36 2831 2830 16:38:15 16:38:15 14:31:35

Should I be looking at the CBYTES & QBYTES columns? When do I know I am about to "run out" of the message queue resources for the present kernel parameter value?

[S.K. Chan · Accepted solution]

I wonder if you can run glance in adviser mode to get kernel utilization information. I did a quick on one of the kernel parameter "nproc", when you run glance and enter "t" it'll show you system table utilization info. I tried with "nproc", this is what I did. In glance the help screen tells me 2 var associated with "nproc"
TBL_PROC_TABLE_AVAIL
TBL_PROC_TABLE_USED
So I created a file called "test" and had this statement in it ..

print TBL_PROC_TABLE_AVAIL, TBL_PROC_TABLE_USED

then I ran glance as such ..

# glance -iterations 1 -adviser_only -syntax test

and I would get 2 column (available and used), you can probably expand on this for your use.

[John Knapp]

S.K. ... Thank you for the info regarding glance-plus ... indeed it does show some kernel parameter status/values .... see below:

System Table Available Used Utilization High(%)
--------------------------------------------------------------------------------
Proc Table (nproc) 2420 345 14 14
File Table (nfile) 4659 1525 33 33
Shared Mem Table (shmmni) 200 16 8 8
Message Table (msgmni) 485 376 78 78
Semaphore Table (semmni) 128 23 18 18
File Locks (nflocks) 200 57 29 29
Psuedo Terminals (npty) 60 8 13 13
Buffer Headers (nbuf) na 202308 na na


Is there a way to get glance-plus to show all configured tunable parameters and their status? Or are the ones available the only ones that make sense in this format?