mmmmn

版权声明：转载时请以超链接形式标明文章原始出处和作者信息及本声明
http://mmmmn.blogbus.com/logs/1427276.html

1、> A few questions regarding NFS 3 in Solaris 9:
>
> o Which ports (both UDP and TCP) do both the client and the

server use?
> o If the number of TCP ports used is small and known, can I

force the
> client and server to use only TCP ports without ill effects?

1) NFS is an RPC service, it's port will almost always change

between
invocations.

2) Well, yes, on your client you may specify the option to

connect via TCP,
however, however performance of TCP is approximately 33%-50%

worse than UDP
NFS (worse with smaller rsize/wsize values).
(mount -F nfs -o vers=3,proto=tcp,rsize=32768,wsize=32768
someserver:/export/example /example).

Typically tcp is only used for NFS over modem links, or other

circumstances
when packet order and integrity can not otherwise be easily

relied upon.

>
>>A few questions regarding NFS 3 in Solaris 9:
>>
>>o Which ports (both UDP and TCP) do both the client and the

server use?
>>o If the number of TCP ports used is small and known, can I

force the
>>client and server to use only TCP ports without ill effects?
>
>
> 1) NFS is an RPC service, it's port will almost always change

between
> invocations.
>
        Is there a range that NFS will use (say, between 2500 and

3500)? So,
if I have the the NFS server behind a firewall which must (long

story)
serve clients in front of the firewall, I would know which port

range I
must keep open?

> 2) Well, yes, on your client you may specify the option to

connect via TCP,
> however, however performance of TCP is approximately 33%-50%

worse than UDP
> NFS (worse with smaller rsize/wsize values).
> (mount -F nfs -o vers=3,proto=tcp,rsize=32768,wsize=32768
> someserver:/export/example /example).
>
> Typically tcp is only used for NFS over modem links, or other

circumstances
> when packet order and integrity can not otherwise be easily

relied upon.
>
        Well, in the long run I would like to run nfs through a

tunnel, so it
would be nice to limit the number of ports I would need. If using

TCP
would do the trick, I am willing to test it out and see what

performance
losses I will be getting. =)

Probably going to Solaris 10 in those machines would also help.

But,
until then...

> Is there a range that NFS will use (say, between 2500 and

3500)? So,
>if I have the the NFS server behind a firewall which must (long

story)
>serve clients in front of the firewall, I would know which port

range I
>must keep open?

The NFS server will generally use 2049 (udp & tcp) for itself;

but there's
statd/lockd and the mount daemon to contend with which makes this
more difficult.

>Probably going to Solaris 10 in those machines would also help.

But,
>until then...

NFSv4 in Solaris 10 or webnfs in older Solaris releases are

restricted
to using a single port in the server.

Casper

2、>I have an A1000 with 2 disks in it. These are in group 1
>and I have the default 10Mb LUN 0 there.

>If I take some disks out of an D1000 and then
>put them in an A1000 they appear in an unassigned
>group in rm6. How do I move them into group 1?

You can use a function in rm6 that expands a drive group.
That won't affect the existing LUNs on that drive group.

>I want to create a LUN using all the spare space including
>the new drives.

Once you've added all of the new drives to the drive group,
you can create a new LUN that occupies all of the newly-added
space. You won't be able to expand existing LUNs. You can
delete the existing LUNs, except LUN 0, and create a new LUN
that occupies all of the remaining space. In that case, you
will have to dump your data before deleting the LUNs, and
restore it afterwards.

Don't delete LUN 0. Raid Manager needs it. Don't even use it
for anything.

As I recall from attempting to add drives to a drive group, it
takes an inordinately long time. It's far easier to delete and
recreate a new drive group. You do need to dump your data before
doing this. However, since this was a number of years ago, I
may be misremembering what it was that took an inordinately long

time.
This was from buying a partially filled A1000 from Sun and adding
drives purchased separately. (Even with an educational discount,

the
Sun drives were significantly more expensive than those from

resellers.)

Except that you can't delete the first drive group because that
contains LUN 0.

I recently replaced all 12 drives in an A1000, upgrading them

from
4 gig to 18 gig drives. I initialized the new drives by plugging
them into an internal slot of an E450 and running

format>analyze>verify
on them, writing over all of the blocks. After doing the firmware
upgrade on the A1000, I powered it down, removed all 12 4-gig

drives,
inserted 4 18-gig drives, and powered it on again. It initialized
the array with one 2-disk RAID-0 drive group with one 10-meg LUN,
leaving the other two drives as unassigned. I was then able to

expand
the drive group up to ten drives, convert it to RAID-5, and

create
two hot spares. The whole procedure didn't take all that long.

--

> Except that you can't delete the first drive group because that
> contains LUN 0.

The A1000s I got all came with some strange "default" setup with

4 luns
in use. SOP was to immediately delete all drive groups and luns

(yes,
even lun 0), then create a hot spare and a big group with a lun

0.

You can certainly delete lun 0 via RM6.

You can certainly delete lun 0 via RM6.

You can, but the RM6.22.1 manuals and release notes warn against
doing that. Previous manuals did advise you to delete all LUNs
and start over. It default configuration seems to have changed
recently as well.

3、> Hi,
>
> When I use the X manager to connect to one of the systems, it

hang on the
> "Starting the Common Desktop Environment CDE Version 1.4" after

supplied the
> login ID and password.
>
> How to solve this problem?
>
> Thanks in advance.
>
>

You can debug it.

ptree `pgrep dtlogin | head -1`

and run "truss -p" on the youngest process ID.

> How to go about it?
>
> # truss -p 6005
> poll(0xFFBEFAA8, 2, -1) (sleeping...)
> signotifywait() (sleeping...)
> lwp_cond_wait(0xFF1234E8, 0xFF1234F8, 0xFF11CD80) (sleeping...)
> lwp_cond_wait(0xFF11CD28, 0xFF11CD10, 0x00000000) (sleeping...)
>
> lwp_cond_wait(0xFF1234E8, 0xFF1234F8, 0xFF11CD80) Err#62 ETIME
> poll(0xFFBEFAA8, 2, -1) (sleeping...)
> signotifywait() (sleeping...)
> lwp_cond_wait(0xFF1234E8, 0xFF1234F8, 0xFF11CD80) (sleeping...)
> lwp_cond_wait(0xFF11CD28, 0xFF11CD10, 0x00000000) (sleeping...)
>
>

It waits for something.
Please post the result of the above ptree command!

 > Here you go:
>
> # ptree -a 9933
> 1 /etc/init -
> 6005 /usr/dt/bin/dtlogin -daemon
> 9933 /usr/dt/bin/dtlogin -daemon
> 9957 /bin/ksh /usr/dt/bin/Xsession
> 9983 /usr/dt/bin/solregis -cd
> 9997 /usr/dt/bin/sdt_shell -c unset _ PWD; .
> /export/home/ope
> 10000 /usr/bin/bash
>

Hmm,

you have likely an offending line in your .profile,
looks like you have "/usr/bin/bash",
which starts another interactive session.

Remove this line!

--

Thanks Michael Tosch.

Problem solved after remove the line.

4、> Hey guys,
>
> Where do I look for to configure my CDE login screens to
> notify the users that their NIS passwords have expired or are

about to
> expire?
>
> Is this a CDE only configuration or other type of password/PAM
> configuration?
NIS map only contains the encrypted password of the shadow file

so the
client machines will not know about password expiartion or

anything
else for that matter.
If using LDAP is out of question for now, you can use a small and

very
effective perl script called "pwaging" and run it on the master

server,
users will receive an e-mail about their password expiration and

get
locked out if ignore mails, something that NIS can not do.
Vahid.