openstack/training-guides - training-guides - OpenDev: Free Software Needs Free Tools

Always no quotes for $() statement.
We don't need quotes to hold blanks in result:
 # i=$(echo 1 2 3)
 # echo $i
 1 2 3
 #
These quotes can make something wrong in some case:
 # i=$(echo '!')
 #
 # i="$(echo '!')"
 -bash: !: event not found
 #
No real problem for current code, only to use a better code style.
Change-Id: Ib31b49680286600c9a182875122a0a752d3f8a33
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com>

It colorizes the output of osbash.sh using
the conventions given in lib-color.sh.
Implements: blueprint osbash-color-support
Closes-Bug: #1443136
Change-Id: I42ca96454fdab47b283222e2deaf3e9b51d03106

Adding orchestration and telementry services require more disk size.
The current base disk size may not be sufficient for newely added
services.
Change-Id: Id8fc90a434cade6e1c9833935938869f6a84252a

With this changeset, the cluster build can be tested starting from a
specific snapshot. This is mostly useful for testing client scripts
without having to go through the whole cluster build every time.
For instance, to rebuild the cluster from snapshot neuntron_configured:
./tools/repeat-test.sh -b -t neutron_configured -s "controller compute"
This will restore the appropriate snapshot on each node, boot nodes
controller and compute, and continue the cluster build from the named
snapshot onward.
Note: this procedure restores, boots, and continues in short order which
uncovers new shortcomings (races) in scripts that assume a service is
available when they should check and wait for the service.
If you just want to restore all nodes to the appropriate snapshots:
./tools/restore-cluster.sh -t <snapshot>
Change-Id: I0b7888d02de6eb50760d421d6d897f516f8c7718

Adding the missing ISO_MD5 variable in the fedora-20-x86_64 lib
Change-Id: I94d90a0373fb8d7b6f57a1252d44e96e76b7e4af
Closes-Bug: #1451434 

Rename scripts in /labs/lib with proper naming convention.
Bug #1451071
Change-Id: I3a1f096414866846a7f9572804c5db852d76a36e

Updates ubuntu server image to ubuntu-server-14.04.2. The older
versions are not currently available for download, resulting in
timeout with http 404 error while download request is being sent.
Change-Id: I86917aa1be0392d45d9e9e33f6a913704ae1fed8

Writing style for bash script is corrected.
Closes-Bug: #1430649
Change-Id: I8c8a73364922d418c47e7cd8ee372710cafab5ce

Adds script with all the colors and decorators like bold, underline.
Uses this script in main osbash.sh script to bring color to the script
run.
Also adds -c option to enable colors while buiding osbash script.
Usage: ./osbash.sh -b cluster (colors enabled by default)
 ./osbash.sh -b --no-color cluster (colors disabled)
Implements: blueprint osbash-color-support
Change-Id: I5b17221f67f76552deb3498b8a020f1b0d3a64a8

This patch makes use of the osbash public and private keys to ssh to VM
rather than the vagrant keys.
This also removes the vagrant keys directory from the repository.
Change-Id: I939274f188a02469c7e2cf9f9502510ea00d514e

This patch replaces the URLs pointing to the
pre-configuration files for all supported distros
to the latest version.
Change-Id: Ia9a158f695562c56a904af36367b3f2acd9db968

This patch adds version 2 of preseed files
pointing to the to the new osbash ssh keys URL
while keeping the older version to point to the
vagrant keys URL.
Change-Id: If0f0b609d5d58d7169eed0e2e84a890c7f3b4970

Currently, we have a file directing the build for each node. However,
rebuilding just one node hardly ever makes sense, so we might as well
remove the extra files and have the user just rebuild the cluster.
With this patch, every distribution will have a basedisk scripts
configuration file (as before) and one scripts configuration file for
building the cluster (or several for building additional, different
clusters).
Change-Id: I7f50d49d7337b790416a2a76c86be510990859fd

Rename config/scripts.<distro> files to config/scripts.<distro>_base,
making it clear that these distro-specific scripts lists are only
used for the base disk.
Change-Id: I358c2df2c8d3f1ddae4fb26850407d2f6fa360ee

Instead of skipping only lines starting with a '#' (comments), also skip
empty lines. For all other lines that are not commands, abort with a
syntax error.
Change-Id: I5e326f2eb27e4aa72ff75a211e36bc3b3a78126f

Mac OS X exports LC_CTYPE=UTF-8 to our ssh environment which makes the
keystone install fail (results in a Python traceback complaining about
"unknown locale: UTF-8").
This patch uses LC_ALL=C to override all locale settings.
The previous, limited work-arounds for neutron are no longer needed and
removed.
Change-Id: Ic7a2f5547fc4bd5d512da5197e48366c83c21c24

This changeset reverts "labs: temporary wbatch fix" and replaces it with
a proper fix, allowing the creation of correct Windows batch files on
build hosts without VirtualBox.
Change-Id: Iad005ce9f0dcecb910fc09b2c43385dd42992eda

This patch automatically removes the corrupt
ISO image if md5sum doesn't match and prompts
user to re-run osbash script as oppose to asking
the user to remove the corrupt image.
Change-Id: I0dae5d7b88980a5644ec684f8cc8f04697f06e8b

The dots printed while waiting for the ssh server to respond are
no longer on a separate line, and the line ends with a newline.
Change-Id: Ib6c1ddaa35a35ff0a1c28f9e98d205740934c0f6

Verify the MD5 checksum of the Linux distribution's ISO image before
trying to use it. Image corruption happens, and the resulting problems
are not always easy to diagnose.
Change-Id: I92852a28392002d4fcd64898453086d5920d0ef1

A new configuration option, BASE_DISK_SIZE, allows users to change the
size of the disk available to the node VMs.
Because the disks grow dynamically, the cost of larger disks is limited
in practice to additional time spent in zero_empty.sh.
Change-Id: I47a72a08cfa51ba6c04306b64e04c4c2600975dd

This patch adds two options to configuration commands (config/scripts.*).
-n <node_name>: change target node of command
-g <gui_type>: override GUI type (gui, headless, sdl) for this boot
The changed syntax is as follows:
cmd boot [-g <gui_type>] [-n <node_name>]
cmd snapshot [-n <node_name>] <snapshot_name>
cmd wait_for_shutdown [-n <node_name>]
cmd snapshot_cycle [-g <gui_type>] [-n <node_name>]
cmd init_node [-n <node_name>]
Change-Id: Id2673c6d504aed33f0259d5751084442e851684b

Remove the old mechanism for queuing scripts.
Change-Id: I7645a80419918b3da585de12802301133fab6d10

Add a new command, queue, to replace the old syntax in config/scripts.*.
Convert all config files to use the new syntax.
Previously, the scripts used hardcoded directory code words: scripts
for $SCRIPTS_DIR, osbash for $SCRIPTS_DIR/osbash.
With the new syntax, the argument is just a relative path inside
$SCRIPTS_DIR. This makes it easier to add (e.g. distribution-specific)
directories.
The new syntax makes also explicit that the command is used to queue
files for later execution.
Change-Id: I414b05785ad322e82f3bf3a38f3374d0e6c45c6c

Exporting via cloned VMs leaves the exported disks and snapshots
registered with VirtualBox (whether or not we pass a "--register" flag
to clonevm). Moving the exported VMs away or trying to export again to
the same location makes VirtualBox complain.
If we register the cloned VM upon creation and then unregister it
afterwards, we get at least rid of the snapshots and have only the
base disk left to unregister manually. This rather complicated method
is implemented here.
The export VMs have "-e" attached to their names, making them easier to
unregister here. They are also easy to recognize when imported into a
VirtualBox that already has node VMs using the original names.
Change-Id: I4532fd3152e705c3b35265359d7e6b0d5facd55e

Spin off get_base_disk_name from get_base_disk_path so we can get the
disk name without the directory.
Change-Id: Id813fd9bbf8a859b639f92a380c067dfc67a95c7

Fix the code responsible for reattaching shared directories after VMs
have been exported.
For OVA export:
- use saved share names instead of default name (for the sake of
 correctness; it is unlikely that a user will change the name)
For dir export:
- save and reattach shared directories after export
Change-Id: I79162c1ec788aa1d89104aa92aebeae2bb05adf5

Make osbashout abort if a client-side scripts returns an error; use a
file named error in the status directory to signal the problem and to
store the name of the offending script for use by Windows batch scripts.
Have osbash.sh and the Windows batch scripts exit when an error file
appears.
Change-Id: I6722e75c9c23b99e5ed25b00429026c2e0ff6a12

Refactoring to make the control flow easier to understand.
Change-Id: Id38fd1c06ba51ec65577e7c6e7d7caf67acff666

With this changeset, osbash host scripts abort installation over ssh if
script execution in the VM returns an error. The patch has no effect on
installations using VirtualBox guest-addons.
Change-Id: Ibc0b297f44289cfe37e07fad1c0fcef1a30abfb2

Move the processing of *.begin status files into a separate function.
Call that function only if the host's status directory is shared with
the client scripts (through VirtualBox guest-additions).
Change-Id: I7716528c02aa4f8b5fc0901ff91ba5d37b5a0d9f

When using the VirtualBox additions, there is a chance that after the
vboxadd service has started, the shared directory is not yet fully
available. In that case, activate_autostart.sh is never run (although
it may appear in time to be deleted).
This patch makes the script from preseed wait for activate_autostart.sh
to become available before it tries to execute it.
Change-Id: I1c2f3ef2f4301f36efec8a70b811ffe97dbaa779

This patch enables host I/O cache for the VMs created by the labs
scripts.
Without host I/O cache, occasional VM filesystem corruption occured on
two out of three host systems when we started taking snapshots after
every service install on the controller VM. Typically, the affected
script would just hang. The issue is somewhat elusive: it occured only
on two out of three test hosts, and the longest observed streak of
correct controller VM builds in a row was 16 (number 17 ended with
filesystem corruption).
With host I/O cache enabled, 125 builds were successful and none failed.
Change-Id: Ib36f44fdbf9366ef781df82f225e5ac8c5bb678f

The recent renaming of local variables missed an instance of a variable.
Fixed.
Change-Id: I75da4382fc35dd288879715d987e8fabb8e2aa47

_vbox_configure_ifs should no longer use the environment variable
NODE_NAME as the name of the node VM. It gets replaced by the correct
value passed as an argument.
Change-Id: I9d36420bcf56bbabe2f47cee2d7eb2ad824bf5d1

If a user changes the permissions of the Vagrant insecure ssh key,
ssh authentication doesn't work anymore and they are prompted for
a password. To prevent that, check (and adjust, if necessary) file
permissions before using ssh.
Change-Id: I90589553de30ef606c4083a1b9980c2cc80a3b3b

Distinguish between the name of a node VM (vm_name) and the hostname
used for the node VM. They just happened to have the same values until
now.
Change-Id: Idc457922fee886ebe504d5f56bbc86d0e5fbbd54

There is no need for users to customize the name of the temporary VM
used for building the base disk. This patch removes that option.
Change-Id: I1e3eebf0c9f1098623d8f4ab50d938211cc78b6f

Make local variables lowercase and some of them more descriptive.
Change-Id: I14072c03243e0fc16aef4a5b6513ca530a99d911

Make local variables lowercase and some of them more descriptive.
Change-Id: I63bbc6a8527864ba3078804eec090a5bfc206f16