FCP_VOLUME_ADDRESSING is supported in snv_113 ?

Finally I have time to install snv_113 to test the fix of the bug 6487899 that I've reported 2.5 years ago. Here's the timetable :

Date Submitted: 2006-10-30 19:45:25 GMT+00:00
Status Changed Date Updated
3-Accepted 2006-11-09 01:58:03
7-Fix in Progress 2009-03-30 09:41:30
8-Fix Available 2009-04-03 04:07:42
10-Fix Delivered 2009-04-22 20:47:53

Oracle buys SUN

SC 3.2 / Solaris 10 u6 / e1000g

Just hit a bug "GLD interfaces unexpectedly send VLAN tagged packets" when luupgraded the SC 3.2 to s10 u6. After the upgrade cluster private interconnects of three T2000 nodes could not come up. Yet another sleepless night at customer's site...

Don't shout at your storage!

Remember: your storage don't like when you shout at it !!!

Brendan, you guys rock!!!

Boot-from-SAN-root-zpool behavior

Root-zpool sitting on SAN disk is configured by default in S10 10/08 with failmode=continue attribute
It means that it...
"Returns EIO to any new write I/O requests
but allows reads to any of the remaining
healthy devices. Any write requests that
have yet to be committed to disk would be
blocked."
The default behavior is: the LUN is offlined 20 sec[1] after the disk disappears from the fabric. The following error message appears on the console:
scsi: WARNING: /pci@780/SUNW,qlc@0,1/fp@0,0/ssd@w5001738000160141,7 (ssd0):
Command failed to complete...Device is gone
when MPxIO is disabled
or
scsi: WARNING: /scsi_vhci/ssd@g0017380000160008 (ssd1):
drive offline
when MPxIO is enabled.

The system is hung and must be restarted.

[1] "Within the LV stack (Leadville), there is a 20-second delay to avoid any
unintentional removal of the cable. This ensures that the loss of sync
is due to a failure and not due to removing the wrong cable and then
realizing the mistake. The 20 seconds is used to correct the mistake.
If they do not put the cable back within 20 seconds, error recovery
will start."

luupgrade problems with second SAS drive on T1000

luupgrade could not configure the second SAS disk of my T1000 as bootable.
It is invisible for the show-disks OBP command...looks like a bug:

{1} ok show-disks
a) /pci@7c0/pci@0/pci@8/scsi@2/disk
q) NO SELECTION
Enter Selection, q to quit: q

probe-scsi shows the disk:

{1} ok probe-scsi
This command may hang the system if a Stop-A or halt command
has been executed.  Please type reset-all to reset the system
before executing this command.
Do you wish to continue? (y/n) y

MPT Version 1.05, Firmware Version 1.10.00.00

Target 0
Unit 0   Disk     FUJITSU MAY2073RCSUN72G 0501    143374738 Blocks, 73 GB
SASAddress 500000e015e69ad2  PhyNum 0
Target 1
Unit 0   Disk     FUJITSU MAY2073RCSUN72G 0501    143374738 Blocks, 73 GB
SASAddress 500000e015e62382  PhyNum 1
<...>

As a workaround, I've shut down the server after unsuccessfull luupgrade and ran this cmd:

{1} ok boot /pci@7c0/pci@0/pci@8/scsi@2/disk@1

It booted fine, but the lustatus is unaware of any boot environments that were created before this boot.

Casablanca

Dialogue no. 1:

Captain Renault:
Chocolate ?
Rick Blaine:
Peanut Butter !

Dialogue no. 2 :

Rick Blaine: GPL v3 ?
Captain Renault: CDDL !

Is no news a good news?

The last news published on OpenSolaris.org site are dated 01/10/2008.
More than 2 months without news...should we care?

Nextra vs Symmetrix

Which storage arrays present themselves as "Array Controller" ?

HP StorageWorks EVA3000/5000 4.004
HP StorageWorks EVA4000/6000/8000
HP StorageWorks XP128/1024
HP StorageWorks XP10000/12000
Compaq Storageworks RA4100

If you know any other storage array presented as "Array controller" (device type 0xC), please let me know.
(Yes, I know about Nextra.)

Update: Symmetrix presents its LUN 0 as "Array Controller" device type when the V-bit (Volume Set Addressing) is turned on.

Solaris Express b81 is released

SXCEb81 aka Nevada build 81, aka snv_81 is available for download.

New packages (compared to b77) :

application SUNWruby18r                      Ruby 1.8.6 patchupdate 110, rubygems 0.9.4 (Root)
application SUNWruby18u                      Ruby 1.8.6 patchupdate 110, rubygems 0.9.4 (Usr)application SUNWstaroffice-sunsearchtoolbar  Sun search toolbar for StarOffice 8
system      SUNWapch22d                      Apache Web Server V2.2 Documentation
system      SUNWapch22m-php5                 PHP5 module for Apache Web Server V2.2
system      SUNWapch22r                      Apache Web Server V2.2 (root)
system      SUNWapch22r-php5                 PHP5 module for Apache Web Server V2.2 (root)
system      SUNWapch22r-php5                 PHP5 module for Apache Web Server V2.2 (root)
system      SUNWapch22u                      Apache Web Server V2.2 (usr)
system      SUNWexpect                       expect - Programmed dialogue with other interactive programs
system      SUNWigb                          Intel 82575 1Gb PCI Express NIC Driver
system      SUNWlibevent                     Libevent - an event notification library
system      SUNWlibtool                      libtool - GNU libtool utility
system      SUNWltdl                         libltdl - GNU libtool dlopen wrapper
system      SUNWmemcached                    memcached (usr)
system      SUNWmemcachedr                   memcached (root)
system      SUNWmemcachedr                   memcached (root)
system      SUNWmysql5r                      MySQL 5.0.45 - MySQL Database Management System (root component)
system      SUNWmysql5test                   MySQL 5.0.45 - MySQL Database Management System (test component)
system      SUNWmysql5u                      MySQL 5.0.45 - MySQL Database Management System (usr component)
system      SUNWndmpr                        NDMP (Root)
system      SUNWndmpu                        NDMP (Usr)
system      SUNWnetcat                       Netcat Commandsystem   
system      SUNWp7zip                        The p7zip compression and archiving utility
system      SUNWphp524-mysql                 PHP5 module for MySQL
system      SUNWphp524-mysql-root            PHP5 module for MySQL(root)
system      SUNWphp524-mysql-root            PHP5 module for MySQL(root)
system      SUNWphp524-pgsql                 PHP5 module for PostgreSQL
system      SUNWphp524-pgsql-root            PHP5 module for PostgreSQL (root)
system      SUNWphp524-pgsql-root            PHP5 module for PostgreSQL (root)
system      SUNWpmdbdpg                      The DBI PostgreSQL Interface for Perl
system      SUNWpmdbi                        Perl Database Independent Interface
system      SUNWservicetagr                  Service Tags (root)
system      SUNWservicetagu                  Service Tags (Usr)
system      SUNWsfe                          SiS900 Ethernet Driver
system      SUNWsmpd                         Target Driver for Serial SCSI Management Protocol (SMP) Compliant Devicessystem      SUNWstosreg                      Service Tags Solaris OS Registration
system      SUNWvim                          Vi IMproved
system      SUNWvncviewer                    VNC viewer client
system      SUNWvscanr                       Virus Scan Service (Root)
system      SUNWvscanu                       Virus Scan Service (Usr)

Removed packages :

system      SUNWglt                          Layout Table Generation Utility
system      SUNWjaxp                         Java API for XML Parsing 1.2
system      SUNWxrgrt                        JAXR Runtime
system      SUNWxrpcrt                       JAX-RPC Runtime
system      SUNWxsrt                         Soap Runtime

Thanks to Cyril for the URL and template!

Put your feedback about the PCA tool here

Martin Paul's PCA tool saved me a lot of time, headache and nerves. That's why I used an opportunity to add my feedback about it in Gerry Haskin's "Patch Corner" blog.

Everybody is welcome do add her 2 cents to the discussion.

X Window cannot start after the fresh install of b77

It's a known bug, the workaround is:

# yes | pkgrm SUNWxwfnt SUNWxwcft SUNWi1of SUNW5xmft SUNWolrte SUNWjxmft SUNWkxmft \
SUNWcxmft SUNWtxfnt SUNWxwoft
# cd <install_dir>/Solaris_11/Product
# yes | pkgadd -d . SUNWxwfnt SUNWxwcft SUNWi1of SUNW5xmft SUNWolrte SUNWjxmft SUNWkxmft \
SUNWcxmft SUNWtxfnt SUNWxwoft
# svcadm restart cde-login

Yes, it's happened

© Rotem Yaari

SunCluster 3.2 in action

Three T2000 servers connected to six racks of decent XIV storage (38TB each) plus FC and GBE switches, that's the HA NFS SunCluster 3.2 in action.

SC3.2 on T2000? Disable the clhbsndr!!!

Useful tip for all those who are running network-intensive services on SunCluster3.2 on T2000/T1000:
Comment the following line in /etc/iu.ap:

e1000g  -1      0       clhbsndr

It will improve the performance.

Why reboot_on_path_failure is useless when you're using ZFS with SunCluster 3.2

Why enabling the reboot_on_path_failure won't help on SC3.2 with ZFS ?
Because the cluster node will panic with

"ZFS: I/O failure"

before the scdpmd will figure out the problem.
What it means is: the ZFS is doing the scdpmd work much faster. It causes the host crash after first IO to the unaccessible (disconnected) storage. The scdpmd is checking storage every 10 minutes (!).
To make it works with ZFS, the bug 6322646 must be fixed.

Here's more detailed explanation.

Enabling jumbo frames for Sun x8 Quad NIC

You need to enable jumbo frames for Sun x8 Express Quad Gigabit Ethernet Adapter?
For Solaris 10 u4 (aka 8/07) these settings won't help :

• ndd -set /dev/nxge0 accept_jumbo 1
• accept_jumbo = 1; in /kernel/drv/nxge.conf

And this one will do the job :

• set nxge:nxge_jumbo_enable=1 in /etc/system
  
• and add "mtu 8000" to /etc/hostname.nxge0

Thanks to Vitaly for pointing me to this workaround.

Join the Pirates !!!

I need to recruit my shipmates to be pirates. It will make me more powerful.
Please join!

Open Windows to see the Sun

Stereoscope on the stereoscopic image

OS Live Upgrade of SunCluster 3.2

We'll start with the good news :

• Solaris 10 u4 is available

• SunCluster 3.2 supports Live Upgrade

What does it mean? Now it's possible to upgrade the OS version of the cluster nodes without interruption of running services. The prerequisites are: the cluster version is 3.2, it's not a single-node cluster and the nodes are running Solaris 10.

Let's do it step-by-step:

1. Create a new BE (Boot Environment)

      lucreate -c s10u3 -m /:/dev/dsk/c0t0d0s3:ufs -n s10u4

2. Upgrade the new BE

      luupgrade -u -n s10u4 -s /net/myserver/s10u4

3. Activate the upgraded BE

      eeprom diag-switch?=false (for SPARC only)

      luactivate s10u4

4. Switchover the Resource Groups and reboot the cluster node

      clrg switch -n node2 rg1

      ...

      clrg switch -n node2 rgN

      init 6

5. Proceed steps 1-3 on the second, third, etc. nodes

Congratilations! We're done:

# cat /etc/release
                   Solaris 10 8/07 s10s_u4wos_12b SPARC
       Copyright 2007 Sun Microsystems, Inc.  All Rights Reserved.
                    Use is subject to license terms.
                        Assembled 16 August 2007

luactivate and EEPROM

luactivate doesn't change the eeprom(1M) boot-device value if the diag-switch? is set to true in PROM

Sun Cluster 3.2 Documentation Center

This page contains links to commonly referenced information in the Sun Cluster 3.2 documentation collection.

People say that:

"In the year 2038, /proc will be so untouchable that systems won't be able to boot" (Bug ID: 6454468)
...in other words:
"The Sun Machine is Coming Down, and We're Gonna Have a Party!!!"
(David Bowie)

JAVAja-ja-ja-ja-ja....

# pkginfo
<...>
system      JAVAjaf                          Java Activation Framework
JAI         JAVAjai-imageio                  Java Advanced Imaging Image I/O Tools
system      JAVAjato                         Java Studio Enterprise Web Application Framework
system      JAVAjatodmo                      Java Studio Enterprise Web Application Framework Sample Application
system      JAVAjatodoc                      Java Studio Enterprise Web Application Framework Sample Documentation
application JAVAjavadb-client                Java DB client
application JAVAjavadb-common                Java DB common files
application JAVAjavadb-core                  Java DB core
application JAVAjavadb-demo                  Java DB demo
application JAVAjavadb-docs                  Java DB documentation
application JAVAjavadb-javadoc               Java DB javadoc
system      JAVAjaxp                         Java API for XML Parsing 1.2
<...>

SC 3.2: first zpool, then quorum

Don't add a configured quorum device to zpool.
Add the quorum after "zpool create/add".

Solaris 10 u4 release date

Solaris 10 u4 will be released in early September.
7/07 that became 8/07 will be 9/07 ?
UPD1: it will stay "8/07"
UPD2: it's based on snv62

How to keep an idiot busy 4 hours

1. Open this link. (update: it doesn' work anymore)
2. Reload
3. Reload
4. ...

SC 3.2 : Which Reservation Protocol should you use - SCSI2 or SCSI3?

In Sun[TM] Cluster Release 3.0 and 3.1, the cluster framework determines which reservation protocol to use based on the configuration. Commencing with Sun[TM] Cluster 3.2 (aka "Solaris[TM] Cluster 3.2") the cluster administrator now has the ability to decide which protocol to use and whether this selection would be on a cluster-wide basis or on a per device basis.

This article gives you some background on the protocols and steps to implement the desired choice. Note that the default SC3.2 behavior still follows the logic of SC3.0 and SC3.1. However, as stated above, you now have the option change the default.

Destroying a zpool under stress causes a hang

We are suffering from this bug and need a fix for it.

Asking the gurus...

http://forum.java.sun.com/thread.jspa?messageID=9791464
http://www.opensolaris.org/jive/thread.jspa?threadID=35909
http://www.opensolaris.org/jive/thread.jspa?threadID=35910

Logo from bash.org.ru

We don't need vodka to be happy...

...having a fresh build with ZFS boot...


I figured out what this picture reminds me...
the Ben Rockwood's blog!

SVM metaset on non-SCSI disks

An SVM metaset creation fails on disks that don't support SCSI-2 reservation with the following error:
"metaset: hostname: setname: failed to reserve any drives"
To avoid this the set must be defined as auto-take metaset:

metaset -s setname -A enable

This automatically prevents SVM from issuing SCSI-2 reservations.