gh-chunks memory allocation problem with 5.1B + Oracle

From: Antonio Gonzalez (antonio.gonzalez@terra.es)
Date: Thu May 05 2005 - 05:37:16 EDT


I wonder if any of you have seen this before:

My environment:
Tru64 v5.1B PK3
Trucluster (2 x GS1280)
Oracle 8.1.7 (single instance)
In the production server (DB1) we have rad_gh_regions up to 4GB for
Oracle SGA.
System RAM totals 8GB

We stop & start Oracle in restricted mode for Export every night. In
this moment, at the Oracle start time, we sometimes get this errors in
the messages file:

May 1 00:06:58 dgttdb1 vmunix: there is currently 0x0 bytes available
in gh-chu
nks memory
May 1 00:06:58 dgttdb1 vmunix: there is currently 0x0 bytes available
in gh-chu
nks memory
May 1 00:06:58 dgttdb1 vmunix: unable to allocate 0x800000 bytes from
gh-chunks
 memory
May 1 00:06:58 dgttdb1 vmunix: unable to allocate 0x800000 bytes from
gh-chunks

Normally this happen a couple of times each month. After two minutes of
erros (with something like 300 errors per second !!) everithing works,
Oracle start normally and the errors stop withut any clue.
We have insert a 2 minutes timeout between the dbshut & dbstart and had
some weeks of peace.

Recently (2 weeks ago) we updated the cluster to PK4. On Monday the
error hit again but this time, after 5 minutes of filling the messages
with 140 MB of error messages (always the same allocation message !!),
everithing freezed, the cluster hung and I needed to power cycle my
boxes to proceed.

So, PK4 have been a real "fiasco". After the patch things become worst
!!.
To my knowledge, the problem is that Oracle doesn't free the memory at
shutdown in an efficient way.

Normally the gh regions work as expected:
db1:/> ipcs -a

Message Queues:
T ID KEY MODE OWNER GROUP CREATOR CGROUP
CBYTES QN
UM QBYTES LSPID LRPID STIME RTIME CTIME
q 0 0x41a2149e --rw------- root system root system
0
 0 16384 524308 524719 12:04:22 12:04:22 9:57:10

Shared Memory:
T ID KEY MODE OWNER GROUP CREATOR CGROUP
NATTCH
            SEGSZ CPID LPID ATIME DTIME CTIME
m 0 0 --rw-r----- oracle oinstall oracle oinstall
38
          8388608 526302 531323 12:15:43 12:13:55 10:02:59
m 1 0 --rw-r----- oracle oinstall oracle oinstall
76
          8388608 526302 531323 12:15:43 12:13:55 10:02:59
m 2 0xcb02cc2c --rw-r----- oracle oinstall oracle oinstall
38
       3858759680 526302 531323 12:15:43 12:15:43 10:02:59

Semaphores:
T ID KEY MODE OWNER GROUP CREATOR CGROUP
NSEMS OT
IME CTIME
s 0 0x696e6974 --ra-r--r-- root system root system
8 9:5
7:10 9:57:10
s 1 0x41a2149e --ra------- root system root system
1 9:5
7:10 9:57:10

Any of you have any advice ??
TIA

Regards from Madrid

Antonio González Ortiz
Consultoría Técnica
e-Mail antonio.gonzalez at terra.es
 



This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:50:18 EDT