complicated shadows

coincidence - confusion - family
today a vivid flashback
Conde Duque, Madrid. 13 July, 1996


romería sin romero

polop de la marina 16 agosto 2014 - 27º - storm threatening

Never done this in a happy state of mind. But made a bad start. Couldn't find any romero. I wanted my own romero. Anyway, sin romero, we set off at a fast pace following no fewer than 6 bands with people interspersed looking decidedly hot; mainly girls in heavy dresses. The pace slowed as we hit the slopes via Plaza Purisima. You have to go back to where you started. And carry romero on a stick. It says it in the rules. So with superstition not on our side, we finally made the 3 hour round trip, spared a thought for the musicians blowing and tapping in the humidity and got back to where we started.


iTunes sync

tolerance quiz

Read the following passage concerning iPad skills. Then answer the simple questions which follow.

A. It hasn't gone into music.
B. What?
A. I can't get it into music.
B. Why not?
A. It says to sign in.
B. Well sign in then.
A. It won't let me.
B. Why?
A. It says try again. I do it and it doesn't let me.
B. How many times have you tried?
A. All day. It says I have to wait 8 hours.
B. Let me have a go.
(pause, keyboard activity)
What was the name of your first pet?
A. What pet?
B. Your first pet. What was its name?
A. I didn't have one.
(pause) Hang on, my Mother had a cat.
B. OK then. What was the cat's name?
A. She called it Busby.
B. No, that's not it.
A. What do you mean it's not it?
B. That's not the name of your first pet. It's not what the iPad thinks it is anyway.
(pause, keyboard activity)
What's your Mother's maiden name?
A. Allen. Why?
(pause, keyboard activity)
B. Is that, A L L A N?
A. I think so.
B. Nope. That's not it either.
A. Yes it is.
B. It's not what the iPad thinks it is.
A. What are you doing?
B. Getting the song onto the iPad.
A. Why? I just want it in music.
B. That's what I'm doing.
A. Why is it asking my pet's name?
B. Are you sure it was Allen?
OK. Look, do you have another e-mail address?
A. Yes.
B. What is is?
A. mary@yahoo.co.uk
B. No, not that one. A different one.
A. I used to have one, but I can't remember it.
B. OK.
(long pause, keyboard activity)
OK. What was the name of your first pet?
A. I didn't have a pet.
B. What about the cat?
A. Busby.
B. Perfect. Now, what is your Mother's maiden name? Allen with an E?
A Yes.
B. Perfect.
(long pause)
OK then. I've changed your password to Creamtea42
A. I'll never remember that.
B. Right. It says that this iTunes account is not valid on this iPad
A. What's iTunes account?
B. It's what I need to get the song onto the iPad.
A. I just want it in music.
B. OK. Look, what do you do when you buy a song?
A. Marie does it.
B. So you've never bought a song.
A. No. Marie does it. She got me that Michael Jackson.
B. OK. Can you call her and ask her for her password? Tell her she can change it afterwards.
A. She says what password are you talking about.
B. (long pause, taking the 'phone)
No, no. Not that one. Your iTunes password. You can always change it afterwards.
She can't remember it.


  1. Estimate a time scale for the above events.
  2. Who set up the iPad in the first instance?
  3. Using your answer to (2) above, or otherwise give:
a. The person to whom we should attribute pet ownership.
b. The person referred to when asked for maiden name.
c. The relationship between character A and Marie.
d. A possible identity for character B.

Your score:
 0-30%, mostly D's: You are good at relationships
30-60%, mostly F's etc. etc.


ubuntu cluster with sssd

winbind is not necessary for ctdb to function correctly. We can leave it sitting there but it does not need to be used for PAM nor nss. So, let's convert our ctdb cluster with winbind to our ctdb cluster with SSSD. The cluster we are converting is this one.

So, on both nodes.

1. Remove the winbind lines from the PAM stack at /etc/pam.d

2. Install sssd
sudo apt-get install sssd
sudo pam-auth-update
choose sssd

3. Create /etc/sssd/sssd.conf
services = nss, pam
config_file_version = 2
domains = altea.site
ad_hostname = smbcluster.altea.site
ad_domain = altea.site
id_provider = ad
access_provider = ad
auth_provider = ad
ldap_id_mapping = false

4. minimise smb.conf
workgroup = ALTEA
realm = ALTEA.SITE
security = ADS
kerberos method = secrets and keytab
netbios name = SMBCLUSTER
clustering = Yes
ctdbd socket = /usr/local/var/run/ctdb/ctdbd.socket
path = /cluster/users
read only = No
path = /cluster/profiles
read only = No
path = /cluster/shared
force create mode = 0770
read only = No

5. create the keytab
remove any old /etc/krb5.keytab
net ads keytab create -UAdministrator

6. edit the kerberos method line in smb.conf
kerberos method = secrets only

7. substitute sss for winbind in /etc/nsswitch.conf
passwd: files sss
group: files sss

7. Create a start-up script (don't laugh)
case $1 in 
echo tearing down apparmor
sudo service apparmor stop
sudo service apparmor teardown
echo startng sssd
sudo rm /var/lib/sss/db/*
sudo service sssd restart
ps aux|grep sssd
echo starting cluster node 1
echo mounting the cluster partition
sudo drbdadm primary r0
mount /dev/drbd1 /cluster
echo loading ctdbd
sudo ctdbd --dbdir=/usr/local/var/lib/ctdb --dbdir-persistent=/usr/local/var/lib/ctdb/persistent --event-script-dir=/usr/local/etc/ctdb/events.d --public-addresses=/usr/local/etc/ctdb/public_addresses --socket=/usr/local/var/run/ctdb/ctdbd.socket --logfile=/var/log/syslog
echo sleeping
sleep 2
echo starting smbd
sudo smbd
sleep 1
echo starting winbindd
sudo winbindd
echo waiting for cluster, tailing log:
sudo tail -f /var/log/syslog
echo stopping cluster
sudo umount /cluster
sudo killall ctdbd smbd winbindd
sudo service sssd stop
ps aux|grep ctdbd
ps aux|grep smbd
ps aux|grep winbindd
echo "done"

8. bring it on
sudo ./ctdbcluster start
node 1
a Lubuntu client


ubuntu samba4 cluster: CTDB ocfs2 DRBD

Mr. Shuttleworth and the packagers at Ubuntu do a great job, but clustering is not one of them. Their samba packages are not built with cluster support and their ctdb is broken. To make matters even worse, the latest samba4 comes without winbind AD support. If that hasn't put you off yet, you then find out that the ctdb source doesn't use standard locations for its directories. Open Source at its very best.

On Ubuntu, clustering means build it yourself. It can be done and when it's up, it's bulletproof. So, here goes. Breathe deeply and. . .

This guide is an ubuntuified version of our openSUSE cluster post. If you think it's missing here, it'll be there instead.

Each node needs a minimum of 2 physical network interface cards, one for the cluster traffic between the nodes and the other that goes out to the domain. Here, we use 2 real cards and a virtual bond for the third.

for domain: altea.site
DC: palmera,
DNS: palmera, bind9_dlz master for altea.site forwarding
node 1: uc1,, and
node 2: uc2,, and

eth0: domain traffic. nic is disabled
eth1: cluster traffic, no route out
bond0: domain and local traffic bound to eth0

cluster disk
You will need a second disk drive for your cluster data. We have a spare disk installed on each of the nodes which is recognised by fdisk as /dev/sdb. We have partitioned it as /dev/sdb1. We have used sdb1 as the shared storage. Under DRBD the cluster device becomes /dev/drbd1. /dev/drbd1 is formatted with ocfs2. We have drbd primaty:primary with both nodes up and mounted.

Get this right and the rest is easy.
/etc/hosts localhost uc1 uc2

Unless you know the ifup syntax for bonding and exactly what you are doing, use the pretty pictures to set up DNS as follows. Here are the screenshots for setting up networkmanager for uc1 on node1. Repeat for uc2 but using the second set of IPs from above.

install the cluster and kerberos stuff
sudo apt-get install drbd8-utils linux-server ocfs2-tools build-essential libattr1-dev krb5-user libblkid-dev gdb libgnutls-dev libreadline-gplv2-dev python-dev autoconf python-dnspython pkg-config pam-krb5utils libpopt-dev apparmor-utils  ldap-utils  libsasl2-modules-gssapi-mit

download and unpack ctdb-2.5.3.tar.gz
then it's just: 
./configure && make
and install it:
sudo make install

configure the cluster
 /usr/local/etc/ctdb/public_addresses bond0 bond0

global {
  usage-count yes;
common {
  net {
    protocol C;
resource r0 {
 net { 
                after-sb-0pri discard-zero-changes;
                after-sb-1pri discard-secondary;
                after-sb-2pri disconnect;
startup { become-primary-on both; }
  on uc1 {
    device    /dev/drbd1;
    disk      /dev/sdb1;
    meta-disk internal;
  on uc2 {
    device    /dev/drbd1;
    disk      /dev/sdb1;
    meta-disk internal;

        ip_port = 7777
        ip_address =
        number = 1
        name = uc1
        cluster = ocfs2
        ip_port = 7777
        ip_address =
        number = 2
        name = uc2
        cluster = ocfs2
        node_count = 2
        name = ocfs2

configure the fs
sudo dpkg-reconfigure ocfs2-tools
Setting cluster stack "o2cb": OK
Starting O2CB cluster ocfs2: OK

partition the spare disk
use fdisk to end up with /dev/sdb1 where:
fdisk -l
gives something like:
Disk /dev/sda: 12.9 GB, 12884901888 bytes, 25165824 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk label type: dos
Disk identifier: 0x000f1bbc

   Device Boot      Start         End      Blocks   Id  System

/dev/sda1            2048     1525759      761856   82  Linux swap / Solaris
/dev/sda2   *     1525760    25165823    11820032   83  Linux

Disk /dev/sdb: 2147 MB, 2147483648 bytes, 4194304 sectors

Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk label type: dos
Disk identifier: 0x000ab2e5

   Device Boot      Start         End      Blocks   Id  System

/dev/sdb1            2048     4194303     2096128   83  Linux

create the drbd metadata on both nodes
sudo drbdadm create-md r0
sudo drbdadm up r0

sync the nodes
Be careful. If you have data you wish to keep on either node then choose that node to begin the sync.
sudo drbdadm primary --force r0
wait for
cat /proc/drbd
to stabilise, then make it active
sudo drbdadm primary r0

format the partition
Unless you just want fail-over, you must use something that doesn't crash if mounted simultaneously.
mkfs -t ocfs2 -N 2 -L ubuntucluster /dev/drbd1

mount it
sudo mkdir /cluster 
sudo mount /dev/drbd1 /cluster
cd /cluster
sudo mkdir users && sudo mkdir profiles && sudo chmod 1777 profiles

download and unpack the latest samba (ATM 4.1.11.tar.gz)
 ./configure --with-cluster-support  --without-ad-dc  --with-shared-modules=idmap_ad && make
install it:
sudo make install

workgroup = ALTEA
netbios name = SMBCLUSTER
realm = ALTEA.SITE
security = ADS
kerberos method = secrets only
winbind enum users = Yes
winbind enum groups = Yes
winbind use default domain = Yes
winbind nss info = rfc2307
idmap config * : backend = tdb
idmap config * : range = 19900-19999
idmap config ALTEA : backend  = ad
idmap config ALTEA : range = 20000-4000000
idmap config ALTEA : schema_mode = rfc2307
clustering = Yes
ctdbd socket = /usr/local/var/run/ctdb/ctdb.socket
path = /cluster/users
read only = No
path = /cluster/profiles
read only = No

create the stack at /etc/pam.d:
auth sufficient pam_winbind.so
auth [success=2 default=ignore] pam_unix.so nullok_secure
auth requisite pam_deny.so
auth required pam_permit.so
auth optional pam_cap.so 
account sufficient      pam_winbind.so
account [success=1 new_authtok_reqd=done default=ignore] pam_unix.so 
account requisite pam_deny.so
account required pam_permit.so
account sufficient pam_localuser.so

create the winbind links:
sudo ln -s /usr/local/samba/lib/libnss_winbind.so.2 /lib/libnss_winbind.so
sudo ln -s /lib/libnss_winbind.so /lib/libnss_winbind.so.2
sudo cp /usr/local/samba/lib/security/pam_winbind.so /lib/security

start the cluster
sudo ctdbd --dbdir=/usr/local/var/lib/ctdb --dbdir-persistent=/usr/local/var/lib/ctdb/persistent --event-script-dir=/usr/local/etc/ctdb/events.d --public-addresses=/usr/local/etc/ctdb/public_addresses --socket=/usr/local/var/run/ctdb/ctdbd.socket --logfile=/var/log/syslog 

(tail the logs and wait for them to stabilise)

 sudo ctdb status
Number of nodes:2
pnn:0     OK (THIS NODE)
hash:0 lmaster:0
Recovery mode:NORMAL (0)
Recovery master:0

create a domain user
uidNumber: 3000016
loginShell: /bin/bash
homeDrive: Z:
homeDirectory: \\smbcluster\users\stevep
profilePath: \\smbcluster\profiles\stevep
unixHomeDirectory: /cluster/users/stevep

'posixify' the group Domain Users
gidNumber: 20513

start smbd
sudo smbd

join the domain
on one node only:
net ads join -UAdministrator

start winbind
sudo winbindd

id stevep
uid=3000016(stevep) gid=20513(domain users) groups=20513(domain users),19903(BUILTIN\users)

add the round robin A records for the cluster to the DC
 host smbcluster
smbcluster.altea.site has address
smbcluster.altea.site has address

 ping smbcluster
PING smbcluster.altea.site ( 56(84) bytes of data.
64 bytes from smbcluster ( icmp_seq=1 ttl=64 time=0.863 ms

sudo chgrp Domain\ Users /cluster/users && sudo chgrp Domain\ Users /cluster/profiles

Linux client:
sudo mount.cifs //smbcluster/users /cluster/users -osec=krb5,username=cifsuser,multiuser
 su stevep
stevep@uc1:/home/steve$ cd ~
stevep@uc1:~$ touch cluster.txt
stevep@uc1:~$ ls -l
total 0
-rw-r--r-- 1 stevep domain users 0 ago 13 10:51 cluster.txt
stevep@uc1:~$ exit

bring up the other node
sudo ctdbd --dbdir=/usr/local/var/lib/ctdb --dbdir-persistent=/usr/local/var/lib/ctdb/persistent --event-script-dir=/usr/local/etc/ctdb/events.d --public-addresses=/usr/local/etc/ctdb/public_addresses --socket=/usr/local/var/run/ctdb/ctdbd.socket --logfile=/var/log/syslog 

sudo drbdadm primary r0

mount /dev/drbd1 /cluster

sudo ctdb enable

sudo ctdb status
Number of nodes:2
pnn:0     OK (THIS NODE)
pnn:1     OK
hash:0 lmaster:0
hash:1 lmaster:1
Recovery mode:NORMAL (0)
Recovery master:0

disable node 1
sudo ctdb disable && sudo ctdb status
Number of nodes:2
pnn:1     OK
hash:0 lmaster:0
hash:1 lmaster:1
Recovery mode:NORMAL (0)
Recovery master:0

(tail the logs and wait for the takeover to finish)

kill smbd on node 1
 killall smbd
 ps aux|grep smbd
steve    18577  0.0  0.1   7124   840 pts/5    S+   11:07   0:00 grep --color=auto smbd
(wait for the log activity to stabilise)

Back to the Linux workstation, check the fail-over:
stevep@uc1:~$ touch cluster2.txt
stevep@uc1:~$ ls -l
total 0
-rw-r--r-- 1 stevep domain users 0 ago 13 11:09 cluster2.txt
-rw-r--r-- 1 stevep domain users 0 ago 13 10:51 cluster.txt
You now admire your work as you fail-over and fail-back time and time again convincing yourself that it does indeed work. At first, you just don't believe it.

More graphical examples below with both Linux and XP clients working on the same data.

Choose your poison: Ubuntu or XP:
all together now, clusters give you. . .
peace of mind. . .


castell l'olla

The nearest I'd ever been with a clear view of the island and the five floating platforms. Slap bang in front the southern platform this year. A good place for the more artistic moments of the show. But all art and design went right out of the window as we approached the climax. The last 5 minutes were deafening and people were grabbing hold fearing the worst as bits of burning firework rained down upon us all on the beach. Spain at its best!


españa: la noche

space station view with (at top) Taurus and The Plaiedes star cluster above the atmosphere



barranc de la canal, polop, 3 de agosto

Adelantada por la sequía, la primera almendra de la temporada.