Tuesday, December 10, 2013

xtrabackup_55: Error writing file 'UNOPENED' (Errcode: 32)

SmartOS and percona-cluster-5.5.31nb3

Error in /var/log/mysql/error.log

WSREP_SST: [INFO] Streaming the backup to joiner at 4444 (20131210 11:16:26.461)
WSREP_SST: [ERROR] innobackupex finished with error: 3.  Check /var/mysql//innobackup.backup.log (20131210 11:16:27.891)
WSREP_SST: [ERROR] Cleanup after exit with status:22 (20131210 11:16:27.912)
131210 11:16:27 [ERROR] WSREP: Failed to read from: wsrep_sst_xtrabackup --role 'donor' --address '' --auth 'sst:secret' --socket '/tmp/mysql.sock' --datadir '/var/mysql/' --defaults-file '/opt/local/etc/my.cnf' --gtid '6ceed930-6185-11e3-ae0d-8e7504b63f54:2'
131210 11:16:27 [ERROR] WSREP: Process completed with error: wsrep_sst_xtrabackup --role 'donor' --address '' --auth 'sst:secret' --socket '/tmp/mysql.sock' --datadir '/var/mysql/' --defaults-file '/opt/local/etc/my.cnf' --gtid '6ceed930-6185-11e3-ae0d-8e7504b63f54:2': 22 (Invalid argument)
131210 11:16:27 [Warning] WSREP: 0 (node1): State transfer to 1 (node2) failed: -1 (Not owner)


And in /var/mysql//innobackup.backup.log you see

xtrabackup_55: Error writing file 'UNOPENED' (Errcode: 32) 

I've removed

streamfmt = xbstream

from /opt/local/etc/my.cnf

Monday, October 14, 2013

Install Baikal server on a SmartOS zone

Baïkal is a Lightweight CalDAV+CardDAV server. The home page is http://baikal-server.com/

Baïkal needs some PHP modules. I use Apache as web server.

pkgin in apache-2.4.6 ap24-php53-5.3.27
pkgin in php53-pdo-5.3.27
pkgin in php53-xmlrpc-5.3.27
pkgin in php53-pdo_sqlite-5.3.27 php53-pdo_mysql-5.3.27
pkgin in php53-dom-5.3.27

Now you must configure php.ini, adding the modules.

vi /opt/local/etc/php.ini


The rest is pretty simple, following the install guide.

Friday, August 9, 2013

MegaCLI on SmartOS

I'm working on a Dell R710 with PERC H700 controller.

# prtconf -v|grep -i raid
                        value='RAID controller'
                        value='MegaRAID SAS 2108 [Liberator]'

# prtconf -v|grep -i perc
                        value='PERC H700 Integrated'

Firts of all go to LSI website http://www.lsi.com/support/Pages/Download-Results.aspx and search by keyword "MegaCLI. Then download Latest MegaCLI (there is Solaris OS).

Unzip it and copy MegaCLI/MegaCli_Solaris/x86/MegaCli to your SmartOS global zone (under /opt/custom/bin for example).

Now edit it with vi (yes, with vi), search for mr_sas and change this word (the driver name) to dr_sas.

Without doing this the utility doesn't detect any controller.

# ./MegaCli -adpCount                              

Controller Count: 0.

Exit Code: 0x00

Thanks to jacques, Nils and elijah on the smartos-discuss mailing list.

Wednesday, August 7, 2013

using send_nsca from SmartOS global zone

If you don't want to install pkgin in the global zone, but you want to use Nagios send_nsca command, this is how I did it.

Inside a zone install the nagios-nsca package.

pkgin in nagios-nsca-2.9.1nb2

From the global zone, cd to the zone zfs and copy these files in a place of your own.

cd /zones/<uuid>

cp opt/local/gcc47/x86_64-sun-solaris2.11/lib/amd64/libgcc_s.so.1 /opt/custom/nsca/

cp opt/local/gcc47/x86_64-sun-solaris2.11/lib/amd64/libgcc_s.so.1

cp opt/local/lib/libmcrypt.so.4

Create a configuration file

vi /opt/custom/nsca/send_nsca.cfg

Now set the LD_LIBRARY_PATH variable and run the command

export LD_LIBRARY_PATH=/opt/custom/nsca; echo "host;passive_check;0;All seems to be OK" | /opt/custom/nsca/send_nsca -H nagios.ip -p 5667 -d ";" -c /opt/custom/nsca/send_nsca.cfg

nsca and send_nsca version incompatibility

I have a Nagios NSCA server running on Debian squeeze. The installed nsca package is the 2.7.2

I want to send passive nagios checks from a SmartOS zone. But the pkgsrc version of nsca client package is 2.9.1

There are incompatibility issues between the 2.7 server 2.7 and a 2.9 client (example http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=670373).

Dropping packet with invalid CRC32 - possibly due to client using wrong password or crypto algorithm?

To resolve the problem without upgrading Debian to a newer release, you can use backports.
Following the instructions I've added a new source to apt

vi /etc/apt/sources.list.d/backports.list

deb http://mi.mirror.garr.it/mirrors/debian-backports squeeze-backports main

apt-get update

And I've installed the newer version of NSCA server.

apt-get -t squeeze-backports install nsca

Now the latest version of send_nsca available on SmartOS works.

Quick testing shows that the 2.7.2 clients seems to work with the 2.9.1 server.

nagios plugin to compare (master/slave) SOA records

This is a Nagios script to compare the master ns and the slave ns SOA records.
So you can know if the slave zone is expired, or is not in sync with the master.



Tuesday, August 6, 2013

mass destroy zfs snapshots

Please pay attention...

zfs list -t snapshot|awk '{print $1}' > /tmp/zfstodestroy

Edit the file to delete the snapshots you don't want to delete!

while read uuid; do zfs destroy $uuid; done < <(cat /tmp/zfstodestroy)

Friday, August 2, 2013

ENOSPC, open '/zones...metadata.json'


Issuing this command

vmadm update f7c4fbb0-aa35-41d4-9d21-614caba785c7 max_physical_memory=2048

I got an error like

ENOSPC, open '/zones/f7c4fbb0-aa35-41d4-9d21-614caba785c7/config/metadata.json'

I try to enter in the zone

zlogin f7c4fbb0-aa35-41d4-9d21-614caba785c7

but I got

[Connected to zone 'f7c4fbb0-aa35-41d4-9d21-614caba785c7' pts/12]
No utmpx entry. You must exec "login" from the lowest level "shell".

[Connection to zone 'f7c4fbb0-aa35-41d4-9d21-614caba785c7' pts/12 closed]


The zone reached the zfs disk quota. Disk full.

But also
vmadm update f7c4fbb0-aa35-41d4-9d21-614caba785c7 quota=50
ENOSPC, open '/zones/f7c4fbb0-aa35-41d4-9d21-614caba785c7/config/metadata.json'

So the solution is simple:

zfs set quota=40G zones/f7c4fbb0-aa35-41d4-9d21-614caba785c7

vmadm update f7c4fbb0-aa35-41d4-9d21-614caba785c7 max_physical_memory=2048
Successfully updated VM f7c4fbb0-aa35-41d4-9d21-614caba785c7

Thursday, August 1, 2013

X11 forwarding request failed on channel 0

I was testing X11 stuff under a SmartOS zone, using the latest pkgsrc-2013Q2, but I got the following error doing an ssh

ssh dest.server -l root -X

X11 forwarding request failed on channel 0

The solution, as suggested by jperkin, is to add this line

XAuthLocation /opt/local/bin/xauth



and restart ssh.

Or you can disable Sun_SSH_1.5 service and install openssh.

two new java tools for dicom files

Remember: I'm not a programmer and I'm not a dicom expert.

I've "released" two java "tools" for dicom files.



This tool (as the dcm2txt tool from dcm4che) read a file and print some dicom attributes: patient name, study date, patientid, studyid, studyiuid.

Or if a directory is passed (instead of a single file) it scan each file and print such attributes.



This tool scan a directory and copy each file to the destination directory creating a path like /destination/study_year/study_month/study_day/patientname_studyi

xclock and rdesktop using a smartos zone

Just for fun


Wednesday, July 31, 2013

bash while loop, function and ssh

Yesterday I struggled an entire morning with a while loop inside a bash script.
The loop ended even if there was a lot of work to do.
And the while loop ended only when I was calling a function inside another function.

The function contained an ssh call to a remote host to get a result.
Such ssh connection, opening a shell on the remote host, maybe killed my while environment?

The solution:

remote() {
  echo -e "Checking if it is the first time we zfs send to remote $1"
  ssh $sshparam zfs list $1 </dev/null &>/dev/null 
  if [ $EL -eq 0 ]
    return 1
    return 0

</dev/null &>/dev/null 

Thursday, May 23, 2013

mirth, smb and netapp filer

Mirth Connect uses jcifs java class.

It seems that there is some problem with authentication on Netapp (ontap) filer.

Thu May 23 15:07:20 CEST [netapp:auth.trace.authenticateUser.loginTraceMsg:info]: AUTH: login from rejected because bad local user password.

Try to add these lines to the Mirth config file mcservice.vmoptions


And restart Mirth

Friday, March 22, 2013

smartos: add additional disk from a different zpool

I have 2 zpools

  pool: zones
  pool: zpool_repos

I want to add to a KVM machine a disk living not in the zones pool, but in the other one.

My VM is already provisioned.
First of all I must stop it.

vmadm stop <uuid>

Now I create a ZFS volume

zfs create -V 10G zpool_repos/mykvmdata

Now I create a json file for the update

vi adddisk.json

   "add_disks": [
          "media": "disk",
          "model": "virtio",
          "nocreate": "true",
          "boot": false,
          "path": "/dev/zvol/rdsk/zpool_repos/uuu"


Now I update the KVM machine.

vmadm update 5942c90f-ecbb-4acd-822f-43e1901e2eb6 -f dataset.json

That's all.
Now, after starting the virtual machine, from inside it, I must partition the new disk (i.e. /dev/vdb in linux) and format.

Thursday, March 14, 2013

iPxe server (on CentOS) to serve SmartOS

This document talks about how I set up an iPXE server on CentOS 6 mainly to serve SmartOS.
Although I've not understood some things (like undionly or the fact that there are many dhcp requests from the pxe client) the following procedure works for me


Install dhcp and tftp services

yum install dhcp.x86_64
yum install tftp-server.x86_64 

Get iPXE undionly

cd /var/lib/tftpboot/
curl http://cuddletech.com/IPXE-100612_undionly.kpxe > undionly.kpxe 
curl http://boot.ipxe.org/undionly.kpxe > undionly.kpxe

Download iPxe

Useful to get some network modules if the client uses gPxe (like KVM one) that is unable to work with iPxe
cd /root
mkdir git
cd git/
git clone git://git.ipxe.org/ipxe.git
cd ipxe/src
make bin/ipxe.pxe
cp bin/ipxe.pxe /var/lib/tftpboot/

DHCP Configuration

It is useful to configure the DHCP server to provide IP addresses only to known MAC addresses. I think it is also useful to link MAC addresses with IP addresses. So a server always get the usual IP.

Note: is the TFTP server (may be the same server as DHCP)

vi /etc/dhcp/dhcpd.conf
ddns-update-style none;

option domain-name "my.domain";
option routers;
option domain-name-servers,;

default-lease-time 600;
max-lease-time 7200;

log-facility local7;

subnet netmask {

   deny unknown-clients;

   host hcn2 {
      hardware ethernet 00:19:99:e0:87:30;
      if exists user-class and option user-class = "iPXE" {
         filename = "menu.ipxe";
      } else {
         filename = "undionly.kpxe";


Configuring iPXE

vi /var/lib/tftpboot/menu.ipxe

######## MAIN MENU ###################
menu Welcome to iPXE's Boot Menu
item --gap -- ------------------------- Operating systems ---------------
item smartos    Boot SmartOS (platform-20121228T011955Z)
item --gap -- ------------------------------ Utilities ------------------
item shell      Enter iPXE shell
item reboot     Reboot
item exit       Exit (boot local disk)
choose --default smartos --timeout 30000 target && goto ${target}

########## UTILITY ITEMS ####################
echo Type exit to get the back to the menu
set menu-timeout 0
goto start

echo Booting failed, dropping to shell
goto shell



########## MENU ITEMS #######################
set base-url ${boot-url}/images/smartos/20130111T010112Z/platform/i86pc
kernel ${base-url}/kernel/amd64/unix -v -B console=text,smartos=true
module ${base-url}/amd64/boot_archive
boot || goto failed
goto start

Preparing the SmartOS image

From the SmartOS wiki:
There are 2 files of importance, the kernel ("unix") and the boot archive ("boot_archive"). The paths to these files are significant and should not be omitted. They should be:
  • (prefix)/platform/i86pc/kernel/amd64/unix
  • (prefix)/platform/i86pc/amd64/boot_archive
The prefix may be anything you wish (a common prefix would be "/tftpboot/smartos/20121004T212912Z/") but the path after the prefix must be in the form above. If you do not the OS will fail to boot with an error saying "krtld: error during initial load/link phase"; in this case the image make sure your path is in the proper form. The important point being, don't just untar the latest platform image into your TFTP directory and boot it, you'll have problems unless you change the typical "platform_20121004T212912Z/" to simply "platform/".


cd /var/lib/tftpboot/
mkdir images
cd images/
mkdir smartos
cd smartos
curl https://download.joyent.com/pub/iso/platform-latest.tgz > /tmp/platform-latest.tgz
cat /tmp/platform-latest.tgz | tar xz
ls | grep platform- | sort | tail -n1
mv platform-20130111T010112Z 20130111T010112Z
cd 20130111T010112Z/
mkdir platform
mv i86pc/ platform/

Please note: change 20130111T010112Z with the current version

Configuring iptables

To allow DHCP

vi /etc/sysconfig/iptables-config

vi /etc/sysconfig/iptables
-A INPUT -m udp -p udp --dport 67:68 --sport 67:68 -j ACCEPT
-A INPUT -m tcp -p tcp --dport 69 -j ACCEPT
-A INPUT -m udp -p udp --dport 69 -j ACCEPT
/sbin/service iptables restart 

Configuring xinetd

To enable TFTP

vi /etc/xinetd.d/tftp

        disable                 = no

/etc/init.d/xinetd restart

Link for today

Solaris command: prstat http://jianmingli.com/wp/?p=1967

Tuesday, February 26, 2013

SmartOS: permanently set VNC port

vmadm update dece98e8-29d7-4394-8cf1-d0185e2258b7 vnc_port=35351

  "vnc": {
    "host": "",
    "port": 35351,
    "display": 29451

Now, you can configure remmina to connect to

Friday, February 22, 2013

smartos vmunbundle exited with code 1

Smartos joyent_20130207T202554Z

vmadm send f59c8669-b709-42cb-98e8-a2f1fa1ee7bf |ssh destination.host vmadm receive

I get this error

vmunbundle exited with code 1

The problem was that, since I was doing some test, there was, on the destination, a zfs dataset for the same zone of a previous attemp.

Solution, on the destination,

zfs destroy zones/f59c8669-b709-42cb-98e8-a2f1fa1ee7bf

Friday, February 15, 2013

SmartOS, kvm, vnc and keyboard

I have had problems with the keyboard using VNC to connect to a KVM guest.
Keyboard mapping I mean.
The solution working for me was:

vmadm update <uuid> qemu_extra_opts="-k it"

linux scp and umask

I use CentOS 5.x
The problem is that if I use scp or sftp to copy a file on such CentOS server, the umask is not taken in consideration.
Behavior of scp and sftp is different.

Btw, a solution is to use openssh 5, but I don't like to alter the distribution, since CentOS 5 install openssh 4.

Using google I've found different solution (pam? etc.) but no one works for me.

So, a way to make what I want is:

For sftp

vi /etc/ssh/sshd_config
# override default of no subsystems
# Subsystem       sftp    /usr/libexec/openssh/sftp-server
Subsystem sftp /bin/sh -c 'umask 0002; /usr/libexec/openssh/sftp-server'

and restart sshd

And for scp

vi /home/myuser/.bashrc

adding (0002 or what you want)

umask 0002

Tuesday, February 5, 2013

smartos update_disks

I want to change the disk model in a KVM machine on SmartOS.

vi upddisk.json

   "update_disks": [
     "path": "/dev/zvol/rdsk/zones/e888a82a-bca1-49fda6d7-988debb09703-disk0",
     "model": "ide"

vmadm update e888a82a-bca1-49fd-a6d7-988debb09703 -f upddisk.json

vmadm get  e888a82a-bca1-49fd-a6d7-988debb09703

Tuesday, January 29, 2013

batch convert dv to avi xvid using ffmpeg

I am using Openindiana.

find bk/Video -type f -exec ~/mmpeg.sh {} \;

The mmpeg.sh script is based on sandipb script.
(Some variables may be unuseful).


The converted files will go in /media/003-9VT166/converted/ directory using as name the full path starting from where find was launched.

So, the file
will become

Please note: I had no time to handle spaces in file names.

dcm4chee query study availability

This query can be useful to list in a range of time the sudies wich availability is ONLINE (0= online, 1= nearline).

SELECT distinct(study.study_iuid)
FROM series
JOIN study ON series.study_fk = study.pk
JOIN patient ON study.patient_fk = patient.pk
JOIN instance ON instance.series_fk = series.pk
where study.study_datetime >= '2000-01-01' and study.study_datetime <= '2010-01-31'
and study.availability = 0 ;

Openindiana: install ffmpeg

pfexec pkg set-publisher -p http://pkg.openindiana.org/sfe
pfexec pkg set-publisher -p http://pkg.openindiana.org/sfe-encumbered
pfexec pkg install ffmpeg

links for today

X11 connection rejected because of wrong authentication after sudo to another user


Monday, January 28, 2013

Another sql query for dcm4chee

Find study instance uid for all the files stored before a date.

SELECT distinct(study.study_iuid)
FROM series
JOIN study ON series.study_fk = study.pk
JOIN patient ON study.patient_fk = patient.pk
JOIN instance ON instance.series_fk = series.pk
JOIN files on files.instance_fk = instance.pk
where files.created_time < '2012-12-15';

Tuesday, January 15, 2013

useful links for today

zone provisioning on admin interface seems to get stuck 

Bad /etc/hosts entry when zone's IP is set to "dhcp"

smartos vmadm failed

Maybe this is only a workaround, like a good craftsman.

I was in this situation:

# vmadm list
UUID                   TYPE  RAM      STATE      ALIAS

76ffffff-cc23-4501...  OS    256      failed     db1

No way to boot or halt the zone.
(Cannot to start vm from state "failed", must be "stopped")
(login allowed only to running zones  ... is 'installed')

The only way to solve the issues, for me, was:

vi /etc/zones/76ffffff-cc23-4501-b5f3-41a69c65b321.xml

and search for a line containing "failed", then delete this line.

Friday, January 11, 2013

today useful links

Trying to chainload iPXE with full feature set from a lesser featured one, whilst still being able to boot non-supported cards with UNDI

Bootstrapping full iPXE native menu with customizable default option with timeout (also includes working Ubuntu 12.04 preseed install)


Monday, January 7, 2013

virtualbox pxe

It seems that Virtualbox uses different PXE "roms" depending on the installation of Extension pack or not.

i.e. without extension pack, it seems to use a limited version of iPXE, and maybe some image format are not handled (Exec format error, Could not boot image, etc.)

Friday, January 4, 2013

Centos Apache and Directory index forbidden

Today I struggled a little bit to make directory index module working.
I'm on Centos.

Installed apache (httpd) rpm.
No changes to configuration. Nothing.

I have created a conf file /etc/httpd/conf.d/my.conf

Alias /images "/var/www/images"

<Directory "/var/www/images">
    Options Indexes MultiViews FollowSymLinks
    AllowOverride All
    Order allow,deny
    Allow from all

I have created the /var/www/images directory.
In this directory I don't want to put any html page. I want to use the Apache autoindex module to serve a list of files contained in this directory.

WTF autoindex not working...

And in the logs /var/log/httpd/error_log you can read

[Thu Jan 03 03:44:05 2013] [error] [client] Directory index forbidden by Options directive: /var/www/images/

The problem was that in /etc/httpd/conf.d/welcome.conf there was

Options -Indexes

- you can delete such file
- you can change in +Indexes

linux: printing pdf in booklet format

I have a 44 pages PDF. I like trees. I want to save money.

So I want to print a booklet (page 1 and 44 on paper 1 side A, page 2 and 43 on paper 1 side B, etc.) from Ubuntu. I can't find the right way to accomplish this from Document Viewer (Evince?).

sudo apt-get install psutils

pdftops -nocrop -paper A4 -expand origin.pdf test.ps

psbook test.ps test-book.ps

psnup -2 test-book.ps test-book2.ps

We use the last command to put 2-pages together.
Note:  /etc/papersize is consulted, so it is better to set a different page size, using psnup -pa4