Thanks for the quick response,
I love this list - it's helped me out of many problematic situations.
Thanks to the following people for their valuable input:
Stuart Davidson
Keith MCCABE
Knut Hellebo
The solution:
1. Under 3.2D you can modify TCP keepalive parameters like
this:(using dbx directly or you could echo commands to dbx)
#!/bin/ksh
dbx -k /vmunix /dev/mem <<EOF > /dev/null 2>&1
assign tcp_keepidle=240
assign tcp_keepintvl=120
assign tcp_maxidle=960
q
EOF
Note: Under ver 4.0 you have the inet subsystem available with these
parameters. All you have to do is stick them in the
/etc/sysconfigtab and reboot.
2. There are apparently tcp/ip patches available for 3.2D and for
3.2G. Also recommended that you upgrade to 3.2G (why am I not
surprised!)
3. Seems like the ndd command may be a SUN command (someone must
have got me really confused...)
4. Indeed, it seems like Sybase as a client server process runs as a
system process and not a user process. The Sybase process -
dataserver was the process that was hogging CPU.
Regarding swap space - perhaps it was absolutely normal that hardly
any swap space was being used but I did find something strange.
#swapon -s shows that I am using /dev/rz25c (4GB) as my swap device.
I noticed in /etc/rc.config that the pagefile was pointing to
/dev/rz24g (RZ28 system disk) which is not being used according to
the fstab. I changed the rc.config to point to /dev/rz25c. And I
modified the swap space allocation method to eager (ln -s /dev/rz25c
/sbin/swapdefault). Soon after that the swap space usage was bumped
up to about 40% - perhaps this was coincedental - After all
/dev/rz25c swap1 is configured in the /etc/fstab - perhaps you also
need to update the /etc/rc.config file for it to take effect. Does
anyone no for sure whether this is true?
Regards
Paulo
Hello osf-managers,
I have a customer running 3.2D-1 Digital Unix (dual CPU - Alpha
8200). They are running a huge Sybase database and have identified a
problem with the tcp/ip for Sybase configuration.
1. There is a tcp/ip parameter "tcp_keepintvl" in the inet subsystem which
needs to be set to approx. 10 minutes. It can be accessed in the following way:
# sysconfig -q inet. They think that this is causing the freezing of
their database ie. a database process finishes but is not killed by
the OS.
However, I noticed that the customer does not have this specific
parameter and subsystem (inet) on his system. It is available for
4.0a which I have verified on my system. What I need to know is - what
is the equivalent of this parameter in 3.2D-1? or what parameter
would directly affect the keep alive interval time of this parameter
(which needs to be set to 10 min)
2. Does anyone know of any specific TCP/IP patches that should be
installed on 3.2D-1? Should I upgrade to 3.2G?
3. Has anyone heard of the # ndd command used to interrogate
tcp/ip parameters? It apparently resides in /usr/sbin. I don't have
it on my system(4.0a)!
4. When I do a swapon -s there is no swapping taking place and yet
the cpu seems to be working quite hard? Is this normal? Would Sybase
as a client server process work in system mode or user mode???
# swapon -s
Swap partition /dev/rz25c:
Allocated space: 523755 pages (4091MB)
In-use space: 2092 pages ( 0%)
Free space: 521663 pages ( 99%)
Total swap allocation:
Allocated space: 523755 pages (4091MB)
In-use space: 2092 pages ( 0%)
Available space: 521663 pages ( 99%)
#vmstat 1 100
Virtual Memory Statistics: (pagesize = 8192)
procs memory pages intr cpu
r w u act free wire fault cow zero react pin pout in sy cs us sy id
6 66 18 109K 6629 11K 768K 34K 330K 147K 115K 57K 358 25K 719 25 42 33
5 67 18 110K 6544 11K 214 26 156 0 25 0 274 3K 348 13 87 0
5 67 18 110K 6544 11K 59 0 55 0 0 0 297 3K 453 10 90 0
5 67 18 110K 6544 11K 55 0 55 0 0 0 218 3K 332 7 93 0
5 67 18 110K 6544 11K 58 0 55 0 0 0 318 3K 487 12 88 0
5 67 18 110K 6544 11K 59 0 55 0 0 0 220 3K 324 8 92 0
5 67 18 110K 6544 11K 57 0 55 0 0 0 210 3K 313 6 94 0
5 67 18 110K 6544 11K 55 0 55 0 0 0 216 3K 329 7 93 0
5 67 18 110K 6544 11K 55 0 55 0 0 0 224 3K 312 5 94 0
5 67 18 110K 6544 11K 56 0 55 0 0 0 202 3K 317 7 93 0
5 67 18 110K 6544 11K 60 0 55 0 0 0 253 3K 376 10 90 0
5 67 18 110K 6544 11K 57 0 55 0 0 0 251 3K 385 7 93 0
5 67 18 110K 6544 11K 56 0 55 0 0 0 237 3K 361 9 91 0
5 67 18 110K 6544 11K 55 0 55 0 0 0 160 3K 240 7 93 0
5 67 18 110K 6544 11K 60 0 55 0 0 0 251 3K 377 7 93 0
5 67 18 110K 6544 11K 62 0 55 0 0 0 307 3K 455 11 89 0
5 67 18 110K 6544 11K 68 0 55 0 0 0 302 3K 468 9 91 0
5 67 18 110K 6544 11K 57 0 55 0 0 0 328 3K 438 11 89 0
5 67 18 110K 6544 11K 56 0 55 0 0 0 216 3K 344 7 93 0
5 67 18 110K 6544 11K 59 0 55 0 0 0 302 3K 451
8 92 0
Thanks, any help to any of the above questions would be appreciated.
I will summarize asap.
Reagards
Paulo
Received on Thu Feb 27 1997 - 08:38:52 NZDT