The cause of some gating failure looks like
because of getting ip address on instance.
However current exercise didn't log the return value.
In this commit, we add get_instance_ip function with
error hanlding support, and apply it on the execise.
Change-Id: I8e17ba68093faafe58a98eb780a032368eea38aa
This is primarily to satisfy the expectations of various scripts that os_*
variables are defined, and will result in a distro similar to "xs6.1.0-59235p"
Fixes bug 1211001
Change-Id: I951e1eb3a5e25f4d8773a11b15cf38157b6492fe
Having --use-mirrors parameter in pip commands causes pip to hang on
some distros.
Pypi uses CDN for long time already, so there's no point to keep this
parameter no more.
Wipe PIP_USE_MIRRORS out of the "function" file.
Change-Id: I70adaf6591834af2482e09eb7f8f9f60df8e7692
Closes-Bug: #1069309
Install a known working recent version of pip that handles installation
dependencies more correctly than before. Extract to a separate script
so it can be used apart from stack.sh.
* Install distro setuptools if it not already present
* Install pip from source tarball as get-pip.py proved to be unreliable
* Remove python-distribute and python-pip from all prereq files,
move python-setuptools to 'general'
* Remove the earlier unfubar_setuptppls() call that attenpted to fix this
* Only update requirements.txt when no changes in repo
Tested on Precise, F18 and CentOS6.
* Fedora and RHEL allow pip to install packages ON TOP OF RPM-installed
packages. THIS IS BROKEN. And is one reason we have to be so picky
about order and so forth.
Change-Id: Ibb4b42119dc2e51577c77bbbbffb110863e5324d
We've agreed that we should be using global requirements, so
force projects to be updated to global requirements before we
test them.
Co-Authored-With: Monty Taylor <mordred@inaugust.com>
Change-Id: I0652f639673e600fd7508a9869ec85f8d5ce4518
We have some complex logic in here to try to do the right things with
the requirements before doing the install of the package which still
winds up being wrong in some cases. Since having written this code,
we've learned that the logic we're trying to achieve is actually what
pip install -e does. So just use that. We have to follow up with a chown
of the resulting egg-info directory, because the sudo command will cause
it to be written by root, which prevents subsequent commands from
operating without privilege in the directory.
Change-Id: Iffd068c94ef84475ebb30758bcf612075d225bea
when ./stack.sh encounters error, the output may look like this:
[ERROR] ./stack.sh:698 nova-api did not start
The source filename is wrong. Actually, it should be like this:
[ERROR] /<path_of_devstack>/lib/nova:698 nova-api did not start
stack.sh uses bash as interpreter, which define "Shell Variables"
BASH_SOURCE:
An array variable whose members are the source filenames where
the corresponding shell function names in the FUNCNAME array
variable are defined. The shell function
${FUNCNAME[$i]} is defined in the file ${BASH_SOURCE[$i]}
and called from ${BASH_SOURCE[$i+1]}.
The function "err" is called by function "die" ( and "err_if_not_set",
and "err_if_not_set" is not used at all). ${BASH_SOURCE[2]} will
ignore the deepest two call entries, which corresponding to the shell
functions: "err" and "die".
In one sentence, this change will print the source filename where the
function is defined and exits via function "die".
Fixes bug #1207658
Change-Id: I2aa6642c5cf4cfe781afe278b3dec3e7cba277fa
Call trace can help user to locate problem quickly.
stack.sh uses bash as interpreter, which defines a series of
"Shell Variables":
BASH_SOURCE:
An array variable whose members are the source filenames
BASH_LINENO:
An array variable whose members are the line numbers in source
files where each corresponding member of FUNCNAME was invoked.
FUNCNAME:
An array variable containing the names of all shell functions
currently in the execution call stack.
run "man bash" and search the variable name to get detailed info.
In function backtrace, it gets the call deepth from
${#BASH_SOURCE[@]}, then print the call stack from top to down.
In function die, backtrace is called with parameter "2" to ignore
the call trace of function "die" and "backtrace".
I add a broken function in lib/database, and call it in stack.sh,
the output looks like this:
[Call Trace]
./stack.sh:104:broken
/home/kui/osd/devstack/lib/database:24:die
[ERROR] ./stack.sh:24 It is broken
Fixes bug # 1207660
Change-Id: I04d0b3ccf783c769e41582c20f48694c19917334
* Adds support for vmdk files as glance images in upload_image function
* Set default image url to use for vsphere driver in stackrc
* Now using a more stable url
Change-Id: If6d45bc8dfd3c812ded5e1bcf69ad7ebd9b64f34
According to the sources the parameter --public in python-glanceclient
should be removed after only using the new parameter --is-public in
Devstack.
Change-Id: I25fbb23f4823b3766db647dd50a5b538aad3e55a
1) Use the newer cell_type configuration option instead of overriding
the compute_api_class.
2) A nova-cells service is started for both the API cell (region) and
the compute cell (child). The screen names were both 'n-cell' which
caused both services to log to the same log file. The screen names have
been changed to n-cell-region and n-cell-child.
3) Security groups are not supported with cells. Multiple exercises are
patched to use the 'default' security group instead of creating a new
one. Additionally, if one uses the 'default' security group, do not try
to delete it, as one can never delete the default.
4) Disable the floating_ips and aggregates exercises when n-cell is
enabled, as they are not supported by cells.
5) Related to #4, disable the floating_ips tests within euca.sh
6) Update the README.md. No services need to be disabled, and one only
needs to enable the n-cell service.
Change-Id: I9782d1e3cda3c9dd3daefa15c043f5b06473cb87
Remove python-netaddr as a DevStack (stack.sh) requirement,
this does not change any individual project requirements.
Specifically it replaces address_in_net() and adds cidr2netmask()
and maskip() functions.
Change-Id: Ic604437fde2e057faced40a310ab282f3eb27726
There is a bug in pip [1] where it will choose to install a package
from an existing build-dir if it exists over the version actually
requested.
Thus if a prior component has installed a later version of the
package, the unpacked code is already in /tmp/$USER-pip-build; it gets
re-installed and manifests in a confusing error along the lines of
---
Downloading/unpacking requests>=1.1,<1.2.3
(from -r /home/stack//python-cinderclient/requirements.txt (line 5))
Running setup.py egg_info for package requests
Requested requests>=1.1,<1.2.3 (from -r
/home/stack/python-cinderclient/requirements.txt (line 5)),
but installing version 1.2.3
...
error: Installed distribution requests 1.2.3 conflicts with
requirement requests>=1.1,<1.2.3
---
I believe pip 1.4 fixes this problem, but it should always be safe to
specify a unique build-directory for pip installs to avoid picking up
old versions.
We also add a cleanup_tmp function for clearing out anything that
stack.sh might leave around when un-stacking, and add a catch-all for
the pip-build dir.
[1] https://github.com/pypa/pip/issues/709
Change-Id: I7ce919cddfd6d6175ae67bd864f82e256ebc7090
Rather than parsing then feeding the pip requirements file to pip,
just have pip consume it directly.
Change-Id: I17bbe4324e6957c7165bc0f340ddae1e51039471
Fedora uses a special PROMPT_COMMAND for screen terminal types,
which are the default with DevStack. The PROMPT_COMMAND interacts
in a negative way with the -t and -X and -p flags, causing DevStack
not to work. To solve this problem, this patch forces PROMPT_COMMAND
to default to /bin/true, triggering no changes to the window title
used within screen. Also this patch enables F19 as a non-FORCE
distribution.
Change-Id: I2414d9e28dd95b69272e132163b29ed83f73b2f6
Fixes: bug #1196340
https://bugs.launchpad.net/devstack/+bug/1068386
"sudo python setup.py develop" installs to /usr/bin on SUSE
Otherwise it fails with "stack.sh:191 g-api did not start".
Change-Id: I1258240ce2a5a765188353fbc8a2e085d0b02fec
Power systems don't have graphics adapters, so we can't simulate a VNC
console. This patch removes that from the default nova configuration
if the system architecture is ppc64.
Change-Id: I129d180b712115e5c275241740d34805fea23e8b
This patch adds two functions to check version strings in the toplevel
functions file. The openvswitch_agent then uses these to compare versions
when checking for tunneling support. The tunneling version check now
also takes into account upstream master versions of Open vSwitch, which
the previous version check always failed on.
Fixes bug #1190734
Change-Id: I0102fb57f8ce5529169025efa21a0996ad68bef1
This is the semi-irregular comment and docs cleanup.
No functional changes should be here although some code is moved in a
small attempt to sort functions and get things where they need to be.
Change-Id: Ib4a3e2590c6fbd016c391acc7aef6421e91c0dca
Simple check to see if EPEL repo is enabled for RHEL6 and, if not,
automated install of the repo RPM.
Also adds an additional sanity check which checking for pip. In the
prior case of EPEL being disabled, a RHEL6 host may not have access to
python-pip. Although this shouldn't happen, its a good sanity check.
Change-Id: I4e8a4dda9475b75d071534d4eef469198502a048
Several guides suggest using data directories under your homedir,
rather than the default /opt area. This is fine, but on RHEL6 and
similar distros homedirs are very restrictive 0700 permissions which
doesn't allow things like httpd to pass through to serve up files.
Even though stack.sh is taking over the host, changing permissions
automatically is not a nice idea. So we just warn when it looks like
this is happening.
Change-Id: I9cd70e7fe90638a2a5c3b8fd94756afacac7c7be
This patch allows DevStack installation on Linux Mint. Installations
on Linux Mint previously failed since its distribution ID did not
match any of the package installation methods and defaulted to using
rpm packages. Linux Mint is a Ubuntu derivative and such does not have
native support for rpm packages.
This change adds "LinuxMint" to the distribution matching options along
with Debian and Ubuntu.
Change-Id: Ice6a201cabe07373a9c1354699777835addeac53
Fixes: bug #1166422
* err() and err_if_not_set() do error-like reporting without aborting the script
* die_if_not_set() now properly dies
* add is_running() from Grenade
Change-Id: I38b88112415a3c07e35bbc2dc65ad839c4d63fce
for files that don't start with a #! or end in .sh, the added tags
are nice for emacs users to automatically switch to the right mode.
Change-Id: If4b93e106191bc744ccad8420cef20e751cdf902
