Using localhost for RABBIT_HOST can cause problems - particularly
on systems where IPv6 is partially broken and localhost resolves
to an IPv6 address.
The service should logically be running on SERVICE_HOST unless
otherwise overridden.
Change-Id: I6efb71b8b95174770066f38403d5d428ec88c3b6
The "-r" option adds io request counts. They are placed next to
the currently present io throughput columns.
Change-Id: I7b4c6f3608afb3714bc997bb4ba005c00eeab61e
Logs are still being cut off in certain circumstances, almost certainly
because the awk process handling the output for verbose logging to a file
dies/is killed early.
Don't kill child processes in the EXIT trap function in this specific case.
Change-Id: I89600b61d59f87e93ff959710d0169e49d2c9845
Ironic setup uses some external scripts that use sudo but they
are not called with a TTY. This patch allows STACK_USER sudo
access without a controlling terminal.
Change-Id: I0c60fcd5d8418738405d4dd9d08fdf5c8f766829
There were a few instances of keystoneclient still being used,
these were replaced with their openstackclient equivalent.
Change-Id: I47fb7cf949b9ab054b97ae387ecbc6d7070fa9c3
It seems that we get a different behavior whether stdout is closed
in the parent process or not when it comes to buffering. If it's
closed we don't get the last line flushed out to the file, if it's
not closed we do. This was seen in the trove pg fail.
A complete shot in the dark, but I'd like to see if this is fixed
with being explicit with printing to the log before stdout.
Change-Id: Ie37299dbc88d549d34d85731e571a5b7ef181311
The logfile output is piped through awk to apply a timestamp
and filter out all of the xtrace commands in the xtrace output.
A while back we added fflush("") which is supposed to flush all
open output files and pipes. It appears that gawk in precise
is old enough that it may only flush stdout, so explicitly
flush the logfile handle.
Change-Id: If5198c2da2a3278eed8ae3d50c7ca5c15eac6d94
If the host ip address is indeterminate while executing
stack.sh, an error message is displayed. This error message
could be a source of confusion since it references localrc,
which is depreciated.
This patch makes the error message clearer and easier to
understand. It does this by taking out the reference to
localrc. It also points the user towards local.conf where there
are suggestions on how to set HOST_IP.
Change-Id: I41f14a2de85449d2a08ab7eb2849844a1087b147
Closes-Bug: #1290556
* Move remaining role creation to create_keystone_accounts()
* Move glance creation to create_glance_accounts()
* Move nova/ec2/s3 creation to create_nova_accounts()
* Move ceilometer creation to create_ceilometer_accounts()
* Move tempest creation to create_tempest_accounts()
* Convert moved code to use OpenStackClient for setup
* files/keystone_data.sh is removed
Note that the SERVICE_TENANT and ADMIN_ROLE lookups in the other service
implementations are not necessary with OSC, all operations can be done
using names rather than requiring IDs.
Change-Id: I4283ca0036ae39fd44ed2eed834b69d78e4f8257
raring EOL was - 27 Jan 2014
f18 EOL was - 14 Jan 2014
opensuse 12.2 was - 15 Jan 2014
if their upstream isn't going to support them, we shouldn't be
in devstack.
this additionally leaves us in an interesting situation that there
is no longer *any* opensuse version listed as supported. if the
opensuse community doesn't step up here we should probably look
at removing it.
Change-Id: Ibb883930b430477dfd3b5126c5db04f95a50d3a7
Move the heat setup which currently happens in files/keystone_data.sh
to lib/heat, where we have create_heat_accounts.
Move the user, role, service and endpoint creation as that is consistent
with what other services, e.g lib/nova are doing.
Change-Id: Iaa2c822cad581d6b2b4f22f8863daf81e25f8485
The various EXTRA_xxx_OPTS variables will be removed in the Juno
development cycle, change the README to reflect the new way for
the Neutron variables.
Change-Id: Ic84da4a9b5a83e66cf0b57d643a87691f15517f0
The bm_poseur git repository link has been broken
for over 11 months. The virtualized/fake baremetal
environment is not working and has not worked in a
long time. Now, on the tail of enabling 'enable -o errexit',
this functionality now has a hard break.
Change-Id: I3cbd8db58c422bc5273d2433278aaa5e449ecfd9
Closes-Bug: 1285954
Some cloud images don't have yum-utils installed, so the call to
yum-config-manager fails. Pre-install it (I still think it's easier
than fiddling config files).
Also, these repo setup steps are using sudo, but the root/sudo checks
happen after this. Move them up before we start trying to do
repo/package installs.
Change-Id: I875e1f0663c9badc00278b2cc1a3b04ca3dde9fc
Check that function calls look like ^function foo {$ in bash8, and fix
all existing failures of that check. Add a note to HACKING.rst
Change-Id: Ic19eecb39e0b20273d1bcd551a42fe400d54e938
devstack should run under -o errexit to ensure that we fail early
when something has gone wrong, otherwise determination of the root
failure location is often quite challenging.
this clears all the normal use cases for devstack, there could be
tests which now die early, which we're going to have to check for
later.
Change-Id: Ibd828c4f4fd95a60d3918d3d7ae90e10649479ab
* Force-flush log output so we don't lose log output in certain error cases.
* Slow down exit paths: add sleep to die(), wait until last moment to
kill child processes (including the awk log output filter)
Change-Id: I1620fd33b89b237d9c2bb6206f3de2c81719f676
Change 4f1fee6eae300a3384900df06ebc857e95854eb0 added the RHEL6
optional repo in fixup_stuff.sh, but it turns out that doesn't get run
until after the package prerequisites phase. Move this into stack.sh
with the RDO repo setup.
Change-Id: Iae0df85fa94c6c1b6f497dd29fda90d03b903a41
Otherwise you get yum errors like [1] when you run stack.sh as !root.
The solution is to run yum commands as root so it can access the right
certs
[1] https://access.redhat.com/site/solutions/312413
Change-Id: I54b0df13508c50aba67e23da11953c536933917a
dstat is far cleaner for getting results out of the environment,
and covers the bulk of our use cases for sysstat and pidstat with
a much better ui.
devstack is allowed to be opinionated, so become opinionated here.
Change-Id: I21ec96339dcd704098512fdafd896738f352962d
we are xtrace happy, however that's just generating bulk in log
files that are mostly ignorable. For the basically bullet proof
functions we should not xtrace.
Change-Id: Iab4e6d270c1546e0db2a06395cefcdf7f7929c3c
support millisecond resolution and the | separator for ts vs.
content.
everything else in openstack is running at millisecond resolution,
and some times it's actually useful to see that when debugging
gate failures.
Change-Id: I2227ab0b4965cd1a24b579bdf2ba8c1f9a432f70
service enabled is something that's not required for sysstat and
friends, because instead we actually can do this with screen_it.
Change-Id: I4aa5787101cb0def46690f38a7f82effbb85f502
Ensure we have a known-good umask. Otherwise files such as
/etc/polkit-1/rules.d/50-libvirt-$STACK_USER.rules may not be readable
by non-root users afterwards.
Also reworded some comments to be more clear.
Change-Id: I7653d4eee062cf32df22aa158da6269b1aa9a558
Closes-Bug: #1265195
The Heat instance-users blueprint requires an additional domain
where heat creates projects and users related to stack resources
so add support for creating this domain when configured to install
Heat. Note a workaround is currently required to make the
openstack command work with the v3 keystone API.
Change-Id: I36157372d85b577952b55481ca5cc42146011a54
migrated most keystoneclient commands from the following libs:
ceilometer
cinder
ironic
keystone
marconi
neutron
nova
savanna
swift
trove
Also need to set and unset openstackclient specific environment
variables from stack.sh
Change-Id: I725f30bc08e1df5a4c5770576c19ad1ddaeb843a
