11-05-2011 04:18 AM
Hello,
On monday I will have my 1st solaris lms installation for many years ago. Because the delivery of the media pack postponed, I need to install from the downloaded lms-4.1-Sol-eval.zip. Im trying to prepare an install-DVD under windows from this image. 7zip complains while extracting about cross overwriting of several equal named files. So I leave this method and burn the image as it is to DVD as ISO9660 Joliet. I hope this problem is filesystem specific.
So I want to ask:
1. contains Solaris 10 an unzip, if not will the unzip v6.0 from http://sunfreeware.com will work?
2. If automounting at Sol10 is not in-the-box, How can you mount an USB disk in Solaris10 CLI in case the DVD will not work?
many thx for your hints, Steffen
Solved! Go to Solution.
11-05-2011 05:14 AM
Recently installed on a solaris10 VM. I was tolled the solaris a default install + patching.
I found the SMTP service was not active and SYSLOG needed additional configuration. The sysadmin turned SMTP on later.
I was able to put the files I had on there with putty's pscp and I could unzip them using the 'unzip' command.
Make sure you can do su or sudo to install. No point in trying if you can't.
Oh yeah, for some reason the installer fails if you try to install from /var/tmp, so don't.
Cheers,
Michel
11-05-2011 03:24 PM
solaris 10 has has zip and unzip bulit in. Just copy the eval to the disk and issue an unzip
as always with solaris, it is good to have the lates patch cluster installed;
also I found a note for a patch concerning the "unzip" command:
https://blogs.oracle.com/patch/entry/need_unzip_fix_available_in
but I am not sure if only patch cluster files do have the problem.
I highly doubt a download to a windows machine, unzip it with windows tools, burn it on DVD and use this as an installation media on solaris. I would not unzip the eval and burn it on DVD, instead, I would burn the lms-4-1-Sol-eval.zip on DVD (or copy it on a USB storage), use this to transfer the file to the server into a new directory other then /var/tmp as Michel pointed out (pscp, ftp, sftp or just insert the DVD in to the drive and copy it to the harddisk :-) ). After the file transfer, again check the MD5 hash. Unzip it and start the installation (as root!).
Usually solaris should be able to automount your USB device, if not try the steps mentioned here:
http://it.toolbox.com/wiki/index.php/Mount_a_USB_key_under_Solaris
11-07-2011 04:44 AM
I forgot to add this link also which also points out other methods:
http://initialprogramload.blogspot.com/2011/02/adding-zfs-zvol-for-extra-swap-space.html
11-07-2011 06:08 AM
on solaris and the appliance the installation log is written to
/var/tmp/Cisco_Prime_install_YYYYMMDD_hhmmss.log
this is a description on how to stop the snmp service and also how to disable it in solaris 10:
11-08-2011 03:34 AM
thanks, for providing the command for syslogd!
you should use /usr/ucb/ps instead of just ps -ef which is /usr/sbin/ps. But you have to execute /usr/ucb/ps -gauxwww as root to get the complete command line for a process; i.e, as root type type the complete path
/usr/ucb/ps -gauxwww | grep -i psu
thats correct with the Activity Log and I also would never start a second psu job before the log entry appears.
11-05-2011 05:14 AM
Recently installed on a solaris10 VM. I was tolled the solaris a default install + patching.
I found the SMTP service was not active and SYSLOG needed additional configuration. The sysadmin turned SMTP on later.
I was able to put the files I had on there with putty's pscp and I could unzip them using the 'unzip' command.
Make sure you can do su or sudo to install. No point in trying if you can't.
Oh yeah, for some reason the installer fails if you try to install from /var/tmp, so don't.
Cheers,
Michel
11-05-2011 03:24 PM
solaris 10 has has zip and unzip bulit in. Just copy the eval to the disk and issue an unzip
as always with solaris, it is good to have the lates patch cluster installed;
also I found a note for a patch concerning the "unzip" command:
https://blogs.oracle.com/patch/entry/need_unzip_fix_available_in
but I am not sure if only patch cluster files do have the problem.
I highly doubt a download to a windows machine, unzip it with windows tools, burn it on DVD and use this as an installation media on solaris. I would not unzip the eval and burn it on DVD, instead, I would burn the lms-4-1-Sol-eval.zip on DVD (or copy it on a USB storage), use this to transfer the file to the server into a new directory other then /var/tmp as Michel pointed out (pscp, ftp, sftp or just insert the DVD in to the drive and copy it to the harddisk :-) ). After the file transfer, again check the MD5 hash. Unzip it and start the installation (as root!).
Usually solaris should be able to automount your USB device, if not try the steps mentioned here:
http://it.toolbox.com/wiki/index.php/Mount_a_USB_key_under_Solaris
11-07-2011 04:25 AM
OK, from this point zip/unzig everithing is OK. But I've got into anothr trouble:
The install routine complains about swap (because of 56Bytes less than recommended) and I need a working procedure to can increase the swap space.
setup.sh has calculated a need of 33480724
# swap -l
swapfile dev swaplo blocks free
/dev/zvol/dsk/rpool/swap 256,2 16 33480688 33480688
# mkfile 8G /dev/zvol/dsk/rpool/swap2
# swap -a /dev/zvol/dsk/rpool/swap2
/dev/zvol/dsk/rpool/swap2: Invalid operation for this filesystem type
# df -h
Filesystem size used avail capacity Mounted on
rpool/ROOT/s10s_u8wos_08a
134G 14G 92G 13% /
/devices 0K 0K 0K 0% /devices
ctfs 0K 0K 0K 0% /system/contract
proc 0K 0K 0K 0% /proc
mnttab 0K 0K 0K 0% /etc/mnttab
swap 17G 424K 17G 1% /etc/svc/volatile
objfs 0K 0K 0K 0% /system/object
sharefs 0K 0K 0K 0% /etc/dfs/sharetab
fd 0K 0K 0K 0% /dev/fd
swap 17G 64K 17G 1% /tmp
swap 17G 40K 17G 1% /var/run
Why does df shows 17GB of swap that should be more than enough?
thx for hints, Steffen
11-07-2011 04:39 AM
have a look at this site which seems to have a good explanation of the steps when using zfs:
http://serverfault.com/questions/17142/how-to-add-swap-to-zfs-disk
11-07-2011 04:44 AM
I forgot to add this link also which also points out other methods:
http://initialprogramload.blogspot.com/2011/02/adding-zfs-zvol-for-extra-swap-space.html
11-07-2011 05:47 AM
thx Martin, its installing now
zfs set volsize=32G rpool/swap
reboot
Two questions yet:
Where do LMS/Sol will place the install.log?
What command disables the system trapd (because installation complains about occopied udp162) permanetly?
Thx, Steffen
11-07-2011 06:08 AM
on solaris and the appliance the installation log is written to
/var/tmp/Cisco_Prime_install_YYYYMMDD_hhmmss.log
this is a description on how to stop the snmp service and also how to disable it in solaris 10:
11-07-2011 06:35 AM
this LMS server is monitored by a different spectrum NMS, so we need to keep SNMP(161) activated, only trapd(162) must be disabled. Is there a different procedure?
I checked also the real behaviour: SNMP-Poll-agent(161) is there but no SNMP-trapd(162). Does it mean the occopied UDP162 was a wrong error message?
# ls /etc/init.d/*snmp*
/etc/init.d/init.snmpdx
# netstat -a | grep 162
*.16236 Idle
*.162 Idle
localhost.33162 localhost.cscoess 49152 0 49152 0 ESTABLISHED
localhost.cscoess localhost.33162 49152 0 49152 0 ESTABLISHED
# netstat -a | grep 9000
#
11-07-2011 07:45 AM
another problem with device update:
Locking the package repository...
Released the Lock...
Installation failed for product [Common Services] with message com.cisco.nm.xms. psu.packagemgmt.InstallerException: Repository in use. Another Package Support U pdater client session may be modifying device support. Locking the package repository...
Released the Lock...
Installation failed for product [Common Services] with message com.cisco.nm.xms. psu.packagemgmt.InstallerException: Repository in use. Another Package Support U pdater client session may be modifying device support.
Where is the PSU lock-file in solaris is located? delete and repeat?
11-07-2011 08:31 AM
solaris can use a special implementation of snmp with a master daemon and registered sub-agenst which is explained here:
https://www-304.ibm.com/support/docview.wss?uid=swg27002148&aid=1
or Net-Snmp as a standalone agent
you should have a look a the man page of snmptrapd on how to change the port. I do not have a system to check to give you any more hint, - sorry.
for the psu isssue, check with
/usr/ucb/ps -auxwww | grep -i psu
if there is still a process running; if yes let it finish, do not kill this process. When it has finished, check if the
/opt/CSCOpx/Psu.Lock
has disappeared. If yes use
/opt/CSCOpx/bin/PSUCli.sh
to do the updates; If not, again check if there is a running psu process, if not you can remove the file.
The error typically comes from the Web-Gui while clickin 2 times the OK (or finish) button...
11-07-2011 10:47 AM
The psu left seams to be from the GUI frist time use for device package update and was crashed after typing the path for the packages and clicking next. I repeated 1st from GUI than from CLI. Than I was searching for the Psu.Lock and after having no idea I rebooted the LMS server maybe 20min from the PSU-GUI crash. Will see tommorow and try your hints.
Do you know Perfromance tuning hints for solaris LMS? Its slower than every windows LMS implementions I previously did.
11-07-2011 10:54 PM
I found the Web-GUI for doing device package updates or SW download very bad in case of reporting the current status of the workflow .
I assume it is the "Finish" button at the end of the workflow. When you click it, it takes a while before a pop-up opens saying that services must be restarted, - here you have to click "OK" and after that, the last page re-appears, still with the "Finish" button enabled and the focus is also set to that button - that's the point where people want to click again "Finish" to start a second process.....
If you are doing CiscoView dev package updates it is not unusual that the workflow takes up to 20 mins.
beside the fact, that the PSUCli tool clearly reports what it is doing, I ve got the feeling that it is much quicker.
I do not know of general steps for performance tuning but there is still the rmeptt (RME performance tuning tool) which can adjust severl parameters for RME tasks:
Performance depends on CPU and RAM, name resolution and other sw running on the system - in virtualized environments the assignment of the resources has influence on the perfomance (static/dynamic assignments) and also I/O can have impact.
on solaris and Linux use
pdshow -stat
to get performance/resource info on every LMS process
11-08-2011 12:31 AM
I agree with you, that GUI is not good for psu and I know also this things with doubleklick of next and you dont know if the next window you expect doesnt appear because you didnt hit the next button correctly or popup blocker, phishing filter and all other kind of this new browser add-on staff had blocked the window. But in our case the GUI crashed completly after klicking next 1st time - murphies law.
I prefere the commandline as well for psu, but if the customers prefere the GUI, than I have to show this is working from GUI.
I heared something about the RME performance tuning tool, but the GUI is slow just after the pure installation without having only one device in the inventory.
Michael told something about of need to adjust the Solaris10 syslogd, Is this true? I didnt found any about in the docomentation.
11-08-2011 01:18 AM
for the GUI to be slow, it also can be AV software on the client side scanning the web browser traffic. If there is AV software enabled on the client side I would try and disabled it for a test; The other things are mainly hardware performance on the server side.
Usually the installation changes the settings for syslog automatically - If you do not have issues then there is nothing to change.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide