Upgrade to Build 100a continued ....

After downloading the CD, booting from it, it fails also to start the Xorg-Server.

Now I booted to single user mode, because booting Build 99 fails because of some ZFS errors
showing shortly and a reboot afterwards.

In Single-User-Mode on Build 100a, I tried to start the Xorg Server on the Fujitsu Lifebook
S-Series with an i855 Graphics card.
The Xorg-Server could not start, because it tries to load an amd64-libpcidata.so and this
is the wron ELF-Class ( 64Bit Class and the Laptop has a 32Bit Processor ).

Now I try to copy ma data back to an external USB disk and the tar command issues some
warnings on files having dates in the future ( 2038-01-01 1:00:00 ) or implausibly old time stamp ( 1970-01-01 01:00:00 )

After this I try to install Build 98 and live till the next build.

OpenSolaris Upgrade to 100a damages the whole system

Today I upgraded to the Build 100a of OpenSolaris with the IPS system. After reboot the system boots
up, but no GUI or command line login appear's.
After power off and reboot to Build 99 the system starts booting but reboots immediatly so the old
version is also not bootable.

The CD with Build 98 boots, but can not import the ZFS pool, because it was opened with an other system.

For now I download the CD of Build 100a and wil lthan try to boot from this CD.

overview tool for CPU/Memory usage in OpenSolaris/Solaris

In my blog-entry from August about the Sun Benchmark Center CD I wrote already
about the tools supplied by the Benchmark Center in Langen/Germany from Sun.
Today I grabbed the CD and did a short look into the contents ( a little late ) and
a found some useful tools here is the Snapshot of cpubar on one of our T1000
server:

You can see each thread on the 6 cores, the process queues, paging, scanrate, memory usage, virtual memory.

On this T1000 the tool runs with about 0.1% CPU usage.

The CD contains tools for SPARC and i386 architecture.
The x86 tools also run on my OpenSolaris (Build 99 ) Laptop.

Installed OpenOffice 3.0 on a Solaris 10 8/07 on SPARC

I have installed OpenOffice 3.0 ( which works fine on my OpenSolaris Laptop ) on
our SunRay-Installation on 2 T1000 servers. Install was fine, but starting OpenOffice lets the process hang:

[ewer@sunray1:/users/trinomic/ewer]>pstack 13677
13677: /opt/openoffice.org3/program/soffice.bin
----------------- lwp# 1 / thread# 1 --------------------
ff345898 lwp_park (0, 0, 0)
fcdc64d4 _XLockDisplay (31008, 31008, fcdc64cc, 2ccc8, 3c550, 31588) + 8
fe553af0 XGetExtensionVersion (31008, fe556e50, 3c550, fe554858, ff36e32c, ff3776f0) + 20
fe556224 _XiCheckExtInit (31008, 1, 3c550, 11e60, fcbe1b30, fe568000) + 8c
fe554858 XListInputDevices (31008, ffbfdfdc, f, ffbfdfe0, fcdc66c0, 0) + 2c
fcbe1b30 _gdk_input_common_init (3c848, 0, 11cc, 1d760, fcb95230, 1000) + 5c
fcbb8fc4 gdk_display_open (0, 3c848, fcc04234, 0, 1, fcbff244) + 2ec
fcb95230 gdk_display_open_default_libgtk_only (0, 6a0a4, fcc96bf0, 0, fcbff244, 3c) + 9c
f9a2a21c gtk_init_check (ffbfe1f4, ffbfe1e4, ff373580, ff272000, 1, 40000000) + 24
fcc96bf0 ???????? (26298, fee061c8, fccf4a5c, 7, 0, 160000)
fcc976c8 ???????? (21c88, 0, fc80bfb8, fcc96998, 26298, fccfb714)
fcc97cf0 create_SalInstance (21c60, fdcb45bc, fccfb6dc, fc816f4c, ffbfe2d8, ffbfe2dc) + 178
fdc09530 ???????? (ffbfe3d8, feda1420, fdca8f70, ffbfe368, 0, fe6f0008)
fdc0a744 ???????? (fdcd0264, 0, fdccf85c, fdc9730d, 1, 1800)
fd9a9ca8 __1cHInitVCL6FrknDcomDsunEstarDunoJReference4n0CElangUXMultiServiceFactory____C_ (ffbfe4d4, 0, fdca8f70, fdcb49fc, 1b68, 1800) + 114
fd9a9800 ???????? (0, ffbfe5a8, fdcb45a8, 0, fdca8f70, fdcb45bc)
fd9a9a00 __1cGSVMain6F_C_ (0, 24310, fdcb45a8, fdca8f70, 2380, fdcb45bc) + 20
fed58138 soffice_main (800, ffbfe5a8, 84c, 800, 848, 800) + c4
00010ab0 ???????? (1, ffbfe704, 0, 0, 0, 0)
00010a94 main (1, ffbfe704, ffbfe70c, 21000, ff2700c0, ff270100) + 10
00010a6c _start (0, 0, 0, 0, 0, 0) + 108
----------------- lwp# 2 / thread# 2 --------------------
ff345898 lwp_park (0, fef7be08, 0)
ff33f850 cond_wait_queue (ff1d8608, ff1d84a0, fef7be08, 0, 0, 0) + 28
ff33fcc8 cond_wait_common (ff1d8608, ff1d84a0, fef7be08, 0, 0, 0) + 298
ff33fe60 _cond_timedwait (ff1d8608, ff1d84a0, fef7bf30, ff06a490, ff3457f8, 0) + 34
ff33ff54 cond_timedwait (ff1d8608, ff1d84a0, fef7bf30, b280e0, 4905688b, 49056881) + 14
ff33ff94 pthread_cond_timedwait (ff1d8608, ff1d84a0, fef7bf30, ff06a490, ff3457f8, fef7bf38) + c
ff02d810 ???????? (a, 59407, ff06a490, b280e0, 4905688b, 49056881)
ff02d994 ???????? (a, 0, ff1d84b8, ff06a490, 5f4, ff1d84b8)
ff3457f8 _lwp_start (0, 0, 0, 0, 0, 0)

It does not matter if I use the German localization or the en-US version.

Installed VirtualBox 2.0.4 on OpenSolaris Build 99

I've installed VirtualBox 2.0.4 on my OpenSolaris Laptop. The installed
OpenSuSE 11.0 worked out of the box.
Sound is still available, but the installed skype does not find a suitable sound device.

activate external monitor connection in OpenSolaris on Lifebook laptop

Some day's ago I did a presentation of SubVersion and Mercurial on my laptop

ewer@solewer:~$ prtdiag | more
System Configuration: FUJITSU SIEMENS LIFEBOOK S6120
BIOS Configuration: Phoenix/FUJITSU Version 1.25 01/29/2004

but didn't get an output signal on the external monitor connector.

So I rebooted to windows, activated the external connector and rebooted
again to OpenSolaris.
The presentation did go on.

This weekend I searched for a solution without the windows reboot cycle.

The first link guided to download a lot of drivers with an install script I installed and uninstalled without testing the drivers.

Today using the hotkey - from the tool before and it worked :-o
So I searched why it worked and found two links 1,2.

The next step would be to used different resolutions and different
dualhead display's, because now I have the same display on both screens.

The tool would be xrandr.

Some help would be found here. For dualhead an xorg.conf is needed, which sets the virtual size to get a dualhead display

Terminal Server Client - applet in OpenSolaris Build 99 for remote GUI access

There is an applet in the GNOME environment which allows to connect to
different systems using the following protocols:

• VNC
• RDP
• RDP5
• XDMCP
• ICA

It allows to store connections and connect via differente protocols.
I just used VNC and RDP for remote connects to different Windows systems.

Firefox 3.0.3 on OpenSolaris Build 99 dumps core during printing with automatic printer discovery

After getting the printers back with the automatic printer discovery in the
service

svc:/network/device-discovery/printers:snmp

They are also listed in the print-Menu of firefox 3.0.3. But Firefox hangs and
a pstack on the process shows a SIGSEGV.

I found a printing solution in the following mailing thread.

For solving the problem it suggests to LD_PRELOAD the libc.so library.

So I inserted into the firefox-Script at the beginning:

LD_PRELOAD=/usr/lib/libc.so
export LD_PRELOAD

After restarting firefox printing worked again.

New Feature of nautilus in OpenSolaris Build SNV_99

Yesterday I red in the Linux User 11/08 ( a German magazine ) about the new
Gnome Version 2.24.

There was written about the enhancements of nautilus, the GNOME file manager.
The "new" Eject-Symbol for removeabel media and that tabs are now supported.
You can open more tabs in one nautilus windows and copy between these tabs.

This is much more an enhancement, because therefore I used the gnome-commander, I posted recently.

Update to automatic printer discovery with printers:snmp

Today I checked the feature of the automatic printer discover with the
service

svc:/network/device-discovery/printers:snmp

which I described in this blog-entry.


The discovery works fine. Get a lot of printers into the panel. The only disadvantage I have currently is, that I have not found a simple GUI way for renaming of the printer names

Found a new filebrowser beside nautilus => Gnome commander

After the last update to Nevada Build 99 of my OpenSolaris system, I found also
the new package
SUNWgnome-commander

It is a filebrowser like the former well known norton commander during the times of DOS.

Replacement of the Heat Sink of CPU0 in a SunFire V240

Last Friday the Service Processor of one of the SunFire V240 mailed a message,
that one of the fan of CPU0 is not running.

System Event Recorded By SC:

HostID: "83549059"

Event: "00040066"

Time: "SAT OCT 18 17:00:58 2008"

Customer Info: "No Entry"

Hostname: "name"

SC IP Address: "xx.xx.xx.xx"

Message: "MAJOR: CPU_FAN @ MB.P0.F1.RS has FAILED."

Logging into LOM also showed that the fan is not running.
After shutting down the server an dismounting the heat sink to for cleaning,
even the processor was pulled out of the socket.

After mounting the heat sink with the processor and power on, the system powered off after some seconds and the LOM issued the message, that there
is no processor in socket 0. So I searched in
the Internet and found the following guide for Sun Fire V240 for replacing the
heat sink.
After a little searching there was also a message asking what to do, when the
processor is on the heat sink, as was the case here.

The only possible thing was to get the processor off from the heat sink with
some force. I took a screw driver and with some caution the processor was
separated from the heat sink.

After putting in the processor into the socket and mounting the heat sink,
the system was running again.

automatic printer discovery in OpenSolaris with printers:snmp failed in Build 98/99

The automatic printer discovery with the SMF service svc:/network/device-discovery/printers:snmp seemed to work, because the service is online.
But when restarting the service again it failed into maintenance.

Here is the error in the log:

[ Okt 16 19:23:20 Executing stop method ("/lib/svc/method/svc-network-discovery stop snmp"). ]
/usr/bin/dbus-send --system --print-reply --dest=org.freedesktop.Hal --type=method_call /org/freedesktop/Hal/devices/network_attached org.freedesktop.Hal.Device.NetworkDiscovery.DisablePrinterScanningViaSNMP
Error org.freedesktop.DBus.Error.NoReply: Did not receive a reply. Possible causes include: the remote application did not send a reply, the message bus security policy blocked the reply, the reply timeout expired, or the network connection was broken.

After some search in the internet, a mailing list thread in opensolaris gave
the hint.

The hal daemon was running in debug-mode and showed, that the SNMP MIP's
were not found and therefore the hal daemon didn't respond. When I remember correct, this is since Build 93.

pfexec pkg install SUNWsmmgr

After installing the package SUNWsmmgr and restarting hal and printer-discovery the service work again.

pfexec svcadm restart svc:/system/hal:default
pfexec svcadm restart svc:/network/device-discovery/printers:snmp

Hope to find the printer's, when I'm back online in the company network.

Updated OpenSolaris - Laptop to SNV_99

Yesterday I upgraded the laptop to the Build SNV_99 with the help of the
IPS package system pkg.

The first problem was, that I had to go over a squid proxy. My fault was the wrong
setting of the HTTP_PROXY environment variable.

The right format is HTTP_PROYX=http://:.
The other thing was to open the port's for the pkg-systems on blastwave ( http://blastwave.network.com:10000/, sunfreeware ( http://pkg.sunfreeware.com:9000/ ) and Life With Solaris ( http://pkg.lifewithsolaris.jp:10000/ - supports a lot of packages for multimedia )

The upgrade was nearly straight forward.
The only problem was, that I installed JDK 1.6.0_07 in /usr/jdk/jdk1.6.0_07.
Pkg tried to install the same version and broke the installation.

Here I had to unmount the new installation and destroy the newly created
boot environment.

beadm umount opensolaris-8
beadm destroy opensolaris-8

After this the next image-update created a new boot environment, but didn't
download all packages newly again.

Now I'm writing from the new installed version.

ewer@solewer:~$ uname -a
SunOS solewer 5.11 snv_99 i86pc i386 i86pc Solaris

Soccer Cake

zpool upgrade in OpenSolaris SNV_98 blusted laptop => grub failed to load stage2

Yesterday, during the day I did a

pfexec zpool upgrade

This showed, that the rpool was on Version 10 and 13 is available.
So I issued

pfexec zpool upgrade rpool

Succeeded without error and I continued to work.
In the evening I shut down the system ( suspend does not work because of
a display@2 which can not suspend ).

At home I started the laptop again and it failed to load stage2 in grub and
I got the grub -shell displayed.

After looking for help in the internet, I found the following discussion on the
OpenSolaris mailinglist.

So I searched for an image of opensolaris with build 98 ( this is the one I'm running ) and found it in the download area of OpenSolaris in the chapter
OpenSolaris 2008.11 Development Builds.
Here is a link to http://genunix.org/ which lists the iso-Images.

I downloaded osol-0811-98.iso, which just supports the primary language,
but the image was 747 MB large and did not fit on the CD, even with overburning.
So I downloaded osol-0811-98-global.iso , which supports all languages
and it was only about 660MB large and fit on the CD.
After downloading I burned the CD.

Today I booted the laptop from the CD.

The first action was

pfexec zpool import

which showed the pool after a

pfexec zpool import -f

got the pool back and than I reinstalled grub

cd /boot/grub
pfexec installgrub stage1 stage2 /dev/rdsk/c3d0s0

The final reboot brought back the boot menu and Windows XP and all the
Solaris Images were available again.

Thank's again, that all is back now. I already thought to have to reinstall
the whole laptop.

automatic backups of a ZFS filesystem

Today I removed accidentally a file on my laptop on a ZFS filesystem. I had a backup on an external disk, but to not trap into that again, I installed now
the ZFS autosnapshot tool from Tim Foster.

I used this valuable tool already on some servers running Solaris 10 on SPARC
hardware.

After downloading you have to extract the archive and call "make"

This generates in the proto subdirectory the package TIMFaut-snapshot.

ewer@solewer:/tmp/zfs-auto-snapshot-0.10/proto$ pfexec pkgadd -d .

The following packages are available:
1 TIMFauto-snapshot ZFS Automatic Snapshot Service
(all) 0.10

Select package(s) you wish to process (or 'all' to process
all packages). (default: all) [?,??,q]: 1

Now you can use a graphical gui and set the backup interval:

/usr/bin/zfs-auto-snapshot-admin.sh simple

You can view your setting also direct in the properties of the ZFS filesystem:

ewer@solewer:/tmp$ zfs get all rpool/export/home
...
rpool/export/home com.sun:auto-snapshot:frequent true local
rpool/export/home com.sun:auto-snapshot:daily false local

Parametrization of the Snapshot ( how many snapshots, how often, .... ) can be
done with svccfg:

ewer@solewer:/tmp$ svccfg
svc:> select auto-snapshot:daily
svc:/system/filesystem/zfs/auto-snapshot:daily> listprop
start method
start/exec astring "/lib/svc/method/zfs-auto-snapshot start"
start/timeout_seconds count 0
start/type astring method
stop method
stop/exec astring "/lib/svc/method/zfs-auto-snapshot stop"
stop/timeout_seconds count 0
stop/type astring method
startd framework
startd/duration astring transient
zfs application
zfs/backup astring none
zfs/backup-lock astring unlocked
zfs/backup-save-cmd astring "not set"
zfs/fs-name astring //
zfs/interval astring days
zfs/label astring daily
zfs/offset astring 0
zfs/period astring 1
zfs/snapshot-children boolean false
zfs/verbose boolean false
zfs/keep astring 14
general framework
general/enabled boolean true
restarter framework NONPERSISTENT
restarter/logfile astring /var/svc/log/system-filesystem-zfs-auto-snapshot:daily.log
restarter/start_pid count 18063
restarter/start_method_timestamp time 1223549190.558374000
restarter/start_method_waitstatus integer 0
restarter/transient_contract count
restarter/auxiliary_state astring none
restarter/next_state astring none
restarter/state astring online
restarter/state_timestamp time 1223549190.566421000
restarter_actions framework NONPERSISTENT
restarter_actions/refresh integer
svc:/system/filesystem/zfs/auto-snapshot:daily>

Here for example I have modified zfs/keep to 14 days instead of the default of 31 days.

Many Thanks to Tim Foster for this superb tool.

convert CVS to Mercurial : abort: decoding near

In my last post I wrote already about the conversion with the version from
the crew-repository and the builtin cvsps usage to convert a CVS
repository to a Mercurial repository:

hg convert --config convert.cvsps=builtin --datesort -A ~/hgUserNames ./gvs ./REPOS/GVS-hg
.....
275 Anpassungen im WFL Bereich für das Thema Abgeltungssteuer
transaction abort!
rollback completed
abort: decoding near 'hristoph K�rner ( ch': 'utf8' codec can't decode bytes in position 11-12: invalid data!

Now on a very large repository I got some aborts about out of memory and
problems in the native cvs executable, but the restart seems to continue.
But the last error as seen above was a conversion problem. Here in Austria
we have some Umlaute which caused problems.

Looking into the mercurial source I found the problem location in
mercurial/util.py in the function "fromlocal" where the UnicodeError is
logged.
The comment above also states the solution to solve the problem:

We attempt to decode strings using the encoding mode set by
HGENCODINGMODE, which defaults to 'strict'. In this mode, unknown
characters will cause an error message. Other modes include
'replace', which replaces unknown characters with a special
Unicode character, and 'ignore', which drops the character.

After I set HGENCODINGMODE the conversion finished.

convert a CVS repository to a Mercurial repostiroy

For evaluation purpose I converted one of our CVS repository to mercurial.

For this purpose I used my OpenSolaris installation which has mercurial
already as package available.

Mercurial is a distributed version control system with no central repository,
which must be available to work with version control.

The conversion of the stable 1.0.2 release of Mercurial needs cvsps in the
path to convert CVS-repositories.

To start, you need the CVSROOT environment variable set to your CVS repository
and a checked out version of the source.

When in .hgrc the convert extension is activated you can start.

hg convert

This will generate a default mercurial repository with the "-hg" appended to
the path.

Here I got a python stack trace with a KeyError: "bad" in cvs.py.
After some looking around, I found a message which says, to use the crew-
version of mercurial ( not the stable one as I found out ), where the external
cvsps is not necessary anymore.

The convert call

Downloads/crew-582dac23ebac/hg -v convert --config convert.cvsps=builtin ./

This converted my CVS repositories without problems.

Start Glassfish as a SMF Service in Solaris 10

I was surprised how easy it was to register Glassfish as a SMF(5) service in
Solaris 10.

The asadmin command already has a target command

bin/asadmin create-service --passwordfile ./passfile domains/domain1

The passfile contains the passwords for the instance:

AS_ADMIN_USER=admin
AS_ADMIN_PASSWORD=
AS_ADMIN_MASTERPASSWORD=

After this, there is a SMF-Service with the name of the domain created.
The service can be started/stopped with svcadm.

Moving Glassfish ressource configuration to a different installation

For a project I configured in the Web-Console ( http://:4848 ) of Glassfish
the JMS and JDBC-Resources necessary for an EJB application.

On a new installation I had to move the Resources.

I installed Glassfish as described on the download site.
Copied the domain.xml in the domain1/config/ directory to the new site.

Here you may have to delete the applications MEjbApp, __ejb_contianer_timer_app and __JWSappclients and the application reference further down in the file.
Just lookup the entries in domain.xml and delete them. These applications
are created dynamically.

You have also to change the the JMS host in the XML-Tag jms-host

After this the new Glassfish has the necessary resources.

standalone Glassfish is slow without netbeans

I used the installed glassfish from the netbeans ide. It worked quite well, but
after some week's I used the asadmin script to start the glassfish without the
netbeans IDE.

The Admin-Console http://:4848 was very slow.
The server log file showed a lot of entries about the Netbeans HTTP-Mointor which is not available.

In <.../domain1>/config/default-web.xml was a filter and mapping configuration for the Netbeans listener.
After deleting the entries and restart of glassfish, the old speed was back
again.