AIX 4.3.3 maintenance release 8 - Any known problems 4

[davemak]

Anybody installed maintenance release 8 for AIX 4.3.3? If so any problems either with install or system problems after installation?

 

[bjverzal]

I've had it running on a few machines for a few weeks. No problems yet. One of them is hit pretty hard on a daily basis. The other is my SP Control Workstation. I haven't done the nodes yet though.

BV

 

[davemak]

BV,

Thanks for info, please post any problems with nodes.

cheers

Dave

 

[risc6k]

Hi davemak
I have experienced ML_08 bug in HACMP
I couldn't correctly conformed, but I guess that ML_08 have got TCP/IP problem in HACMP.
sorry for bed english
hope to advice.
thanks & regards

 

[DataJanitor]

I installed it last weekend. The only nuisance I noticed is that for some reason, now I can't start Service Director from the X version of SMIT. I probably need to upgrade something else.

 

[aixqueen]

netscape also breaks at 4.3.3. mainenance 8
Netscape is broken after upgrade to 4.3.3-08


exec(): 0509-036 Cannot load program
/usr/netscape/communicator/us/netscape_aix4 because of the
following errors: 0509-130 Symbol resolution failed for
/usr/lib/libpthreads.a(shr.o) because: 0509-136 Symbol
thread_unlock (number 123) is not exported from dependent
module /usr/netscape/communicator/lib433/libc_r.a(shr.o).
0509-136 Symbol thread_waitlock (number 124) is not exported
from dependent module
/usr/netscape/communicator/lib433/libc_r.a(shr.o). 0509-192
Examine .loader section symbols with the 'dump -Tv' command.

The problem is that the libc.a that comes with Netscape is not functioning.
One temporary fix to get it running:

You need to copy /usr/lib/libc.a over /lib433/libc.a or make a link from there to
the system lib. Netscape's copy of libc.a is not working with ML-08, just the
system one is needed.

 

[bi]

You can upgrade netscape to 4.67i and the problem is fixed. Here is what I got from IBM:

ftp://aix.software.ibm.com/aix/efixes/netscape/aix43_installp

You will need to download three filesets:

Netscape.communicator-us
Netscape.help.en_US.communicator
Netscape.msg.en_US.communicator

Then, of course, use installp or smitty to install the filesets.

 

[DataJanitor]

Thanks for the netscape heads-up. I hadn't checked that and didn't realize it wasn't working.

Meanwhile, I found another peculiar thing. Although FTP appears to be working, I'm getting syslog error messages since the upgrade saying:

ftpd[24254]: Failed dlopen: /usr/lib/libpag.a(shr.o): A file or directory in the path name does not exist.

Any clues?

 

[aixqueen]

IY18732 - (AIXV43 only) FTPD LOGS DLOPEN ERROR

Filesets needed for selected item Information file Byte Size
bos.net.tcp.client.4.3.3.54
bos.rte.tty.4.3.3.52

 

[jmss]

I'm currently investigating a problem with a P680 ( 6 way / 8Gb RAM ) running Oracle ( 8.1.7 ), and AIX (4.3.3 Maint Level 8 ), which does not appear to be managing its memory correctly - it appears to be holding large amounts of data in memory even if Oracle and its associated applications are shutdown,

This degrades performance over time as it holds betweem 2/4 Gb each time Oracle is started / shutdown.

Has anyone else experienced anything like this ?

 

[raylin]

Hi
Is the P680 a 64bit machine ?
What version of 8.1.7 have you installed : 32 or 64 bit ?

 

[raylin]

Hi
Is the P680 a 64bit machine ?
What version of 8.1.7 have you installed : 32 or 64 bit ?

 

[jmss]

The P680 is a 64Bit machine, but the kernel is 32bit and uses kernel extensions.

We have installed both 32bit and 64bit 8.1.7 Oracle to try to resolve this issue and the only difference is that the 32bit uses the memory slower than the 64bit - but the end game is the same - it holds onto the memory.

I'm just rebooting the P680 and am not going to start Oracle as over the weekend All the memory got used whilst a mksysb was running - dont know why yet !

If anyone could run a vmstat 2 - and see how much is in the fre column - does it ever go up ?

If anyone runs a svmon -S -t 10 and see which segments are being used in memory - how big are they ( 4k pages )

Any ideas will be helpful

 

[raylin]

In the moment I have installation problems with 64bit 8.1.7 ...
Question ?
Could you "connect internal" without be prompted for a password with 64bit 8.1.7 Oracle ?
I have found a workaround for this password prompting in a "pure oracle environment" with the init.ora parameter REMOTE_LOGIN_PASSWORDFILE=EXCLUSIVE and the orapw command ...but I am trying a SAP installation with 8.1.7 64bit but the SAP installation scripts seems to connect internal without a password and it is difficult to hack the setup script ...
But I will look at a 32bit 8.1.7 installation to see if I have too a memory "cleaning" problem

 

[raylin]

Here the output from vmstat and svmon aix4.3.3 ml 08 32bit 8.1.7 oracle . There is no heavy load . Its is a newly installed SAP fi/co with Oracle on a SP node with 2GB ram 2 power3 375 MHz :

root@bssp35a0:/home/root>vmstat 2 5
kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
0 0 349349 6674 0 0 0 13 27 0 133 106 68 8 13 78 1
1 1 310204 45824 0 0 0 0 0 0 337 80805 129 32 22 34 12
0 1 310204 45824 0 0 0 0 0 0 243 285 84 6 0 94 0
0 1 310204 45824 0 0 0 0 0 0 243 136 88 0 1 99 0
0 1 310204 45824 0 0 0 0 0 0 245 133 82 0 1 99 0
root@bssp35a0:/home/root>svmon -S -t 10

Vsid Esid Type Description Inuse Pin Pgsp Virtual Addr Range
d224 - pers /dev/sapdata6_lv:341 57043 0 - - 0..124999
b727 - work 39721 1 0 39714 0..39841 :
64595..65535
3b0df - work 38592 0 10 38602 0..65535
233f3 - work 29377 0 5 29382 0..40848
18eae - work 18457 0 5 18462 0..65535
26db1 - pers /dev/sapdata1_lv:205 17680 0 - - 0..153601
1f5ad - work 17400 0 0 17400 0..32451
22171 - work 10925 1 60 11044 0..10967 :
65310..65535
38dde - pers /dev/sapdata1_lv:614 9332 0 - - 0..281601
384bc - work 9301 1 16300 18504 0..18467 :
65273..65535

 

[jmss]

So if you stop your application, does the value in the FRE column return to a reasonable level - IE does it free more than just the Oracle SGA.

If you then do a svmon again - have you still got these areas of memory still allocated - we have !!

Thanks for your help with this - it is most useful

 

[raylin]

values after stopping SAP processes and DB processes :

root@bssp35a0:/home/root>vmstat 2 10
kthr memory page faults cpu
----- ----------- ------------------------ ------------ -----------
r b avm fre re pi po fr sr cy in sy cs us sy id wa
0 0 117927 237521 0 0 0 13 27 0 133 171 68 8 13 78 1
0 1 117927 237520 0 0 0 0 0 0 244 613 80 0 0 99 0
0 1 117927 237520 0 0 0 0 0 0 247 135 78 0 0 99 0
0 1 117927 237520 0 0 0 0 0 0 236 139 81 0 0 99 0
0 1 117927 237520 0 0 0 0 0 0 235 128 79 0 0 99 0
0 1 130446 224997 0 0 0 0 0 0 234 68871 72 22 28 50 0
2 1 158182 197259 0 0 0 0 0 0 230 242415 74 25 60 15 0
1 1 117925 237522 0 0 0 0 0 0 244 147556 74 24 26 50 0
1 1 155325 200190 0 0 0 0 0 0 232 208836 74 24 72 5 0
1 1 144082 210830 0 0 0 0 0 0 345 249009 125 16 40 37 7
root@bssp35a0:/home/root>svmon -S -t 10

Vsid Esid Type Description Inuse Pin Pgsp Virtual Addr Range
d224 - pers /dev/sapdata6_lv:341 57049 0 - - 0..124999
1974e - work 31777 0 0 31771 0..40747 :
64595..65535
b727 - work 30142 2 0 30142 0..40747 :
64595..65535
26db1 - pers /dev/sapdata1_lv:205 17682 0 - - 0..153601
22171 - work 10925 1 60 11044 0..10967 :
65310..65535
38dde - pers /dev/sapdata1_lv:614 9332 0 - - 0..281601
384bc - work 9302 1 16300 18505 0..18468 :
65273..65535
111aa - pers /dev/sapdata6_lv:270 8192 0 - - 0..122879
1582 - work 8075 2 0 8075 0..40747 :
64595..65535
c006 - work page frame table 7168 7168 0 0 0..7167

 

[jmss]

Raylin,

Thanks very much for you're help - this is what I would expect - unfortunately this does not reflect what is happenning on our box - it only returns the SGA ( about 300 Mb )

Is there any change you could details your vmtune / schedtune details.

Do you use asynchronous i/o ?

Can you think of anything else

 

[raylin]

yes I use asynchronous i/o . i have not done any tuning work in the moment. output of kernel extension :

root@bssp35a0:/home/root>/etc/loadext -r

Oracle Kernel Extension Loader for AIX
Copyright (c) 1998,1999 Oracle Corporation

> >
Currently running Kernel Extension Path: /etc/pw-syscall
Kernel Extension Version: 3


Did you experience the password prompting with 64bit 8.1.7 (connect internal) ?

 

[jmss]

raylin

When connecting to your database try doing an :

sqlplus ' / as sysdba '

which implicitly connects to the database an removes the need to connect internal


BTW - my loadext returns the same as your's

if you do a lsattr -E -l aio0 , what do you get