The comment indicates that tox is used to generate config files, but
it's actually not.
Change-Id: Iddd45e084641bb3711985484e988ed533965f4f4
Signed-off-by: Takashi Kajinami <kajinamit@oss.nttdata.com>
Isolate dependencies for generate-devstack-plugins-list.sh by
creating a dedicated Python virtual environment with required
packages. This ensures the script uses the correct version of
dependencies regardless of system Python configuration.
Closes-Bug: #2138869
Change-Id: I8a7e49b0b908aec7c95b7fbfb6c9a16375f267a6
Signed-off-by: Abhishek Bongale <abhishekbongale@outlook.com>
The job stopped working recently, the reason is unclear, but since Jammy
is no longer included as required platform, let's just stop testing it.
Signed-off-by: Dr. Jens Harbott <harbott@osism.tech>
Change-Id: I796eddac96313584f4aa5225a9b9803bb2ad19b5
The stop_ovn function stops services but leaves behind runtime files
and stale configuration that can cause restacking failures. This change
improves robustness by:
- Making _stop_process check is-active in addition to is-enabled, so
services are stopped even if they are running but not enabled.
- Clearing OVS external-ids before stopping to prevent stale config
(ovn-remote, ovn-bridge, etc.) from persisting across restacks.
- Cleaning up runtime files (*.sock, *.pid, *.ctl) in both OVS_RUNDIR
and OVN_RUNDIR after stopping services, as stale sockets can prevent
ovsdb-server from binding on restart.
- Removing database lock files (.*.db.~lock~) which can block database
access if services crash or stop uncleanly.
These changes allow stack.sh to run successfully after unstack.sh
without requiring a full clean.sh.
Generated-By: Cursor claude-opus-4.5
Change-Id: I8736f19a8892200948ee74854f99fd99eed5110b
Signed-off-by: Sean Mooney <work@seanmooney.info>
This change refactors the MySQL/MariaDB configuration to:
- Allow both unix_socket and mysql_native_password authentication
using the MariaDB 'IDENTIFIED VIA ... OR ...' syntax. This enables
restacking without needing to reset authentication in unstack.sh.
- Add use_mariadb_socket_auth helper variable to simplify the complex
conditional logic (addresses TODO comment).
- Fix missing DATABASE_USER@'%' creation for modern Debian/Ubuntu
with MariaDB socket auth.
- Fix inconsistent distro checks that were missing trixie.
- Remove dead Oracle Linux code since it's not in SUPPORTED_DISTROS.
Oracle Linux is still handled as RHEL-like via is_fedora().
Generated-By: Cursor claude-opus-4.5
Change-Id: I4becbfe6325bcb29deef8e50e9a9f05678f47802
Signed-off-by: Sean Mooney <work@seanmooney.info>
Several services started by stack.sh were not properly stopped in
unstack.sh, which could leave processes running and prevent clean
restacking:
- tcpdump: lib/tcpdump was not sourced, so stop_tcpdump was unavailable.
Added source and stop call when tcpdump service is enabled.
- s-container-sync: Swift's container-sync daemon was started via
run_process but not explicitly stopped. Added stop_process call in
stop_swift function.
- ldap: The stop_ldap function existed but was never called. Added
stop call in UNSTACK_ALL section, consistent with mysql/rabbit
handling.
These fixes ensure all services started by stack.sh are properly
stopped by unstack.sh, allowing clean restacking.
Generated-By: Cursor claude-opus-4.5
Change-Id: I8bcc6fe82264bb35a616dae39f4216ba6200b547
Signed-off-by: Sean Mooney <work@seanmooney.info>
Neutron services (neutron-api, neutron-rpc-server, and
neutron-periodic-workers) were not being stopped during unstack
because they were not detected as enabled in ENABLED_SERVICES.
The root cause was that these services were dynamically enabled
during stack.sh execution via inline enable_service calls in
start_neutron_service_and_check(), but this logic was not
replicated in unstack.sh. When unstack.sh called stop_process
for these services, the is_service_enabled check failed because
the services were not in ENABLED_SERVICES.
This fix creates a shared enable_neutron_server_services()
function that encapsulates the service enabling logic, reading
neutron.conf to determine which services should be enabled
(including conditional RPC worker enablement based on the
rpc_workers configuration). This function is now called from
both the stack.sh path (in start_neutron_service_and_check)
and the unstack.sh path (after loading plugin settings).
This ensures both stack and unstack use identical logic to
determine enabled services, allowing stop_process to properly
detect and stop all neutron server services.
Generated-By: Cursor claude-sonnet-4.5
Change-Id: I6179f3a861401ff12178aaee8b82ba7bf71dd765
Signed-off-by: Sean Mooney <work@seanmooney.info>
An error was introduced in [0], it added an incorrect dependency on
genisoimage for nova, and kept a valid dependency on xorriso in n-cpu.
It breaks centos 10 deployments, the expected behavior is to use xorriso
on modern rhel/centos.
[0] https://review.opendev.org/c/openstack/devstack/+/963799
Change-Id: I0057f934c453ded198f8b5e58f4924260a3d2508
Signed-off-by: Gregory Thiemonge <gthiemon@redhat.com>
The file should be owned by the user(STACK_USER) to run the process.
Note that STACK_USER may not match the current user in case stack.sh is
run by root.
Also we should not assume that the group name always matches the user
name.
Change-Id: Iec300311cab1b1a2fa124879aa3dc6a57a6a706b
Signed-off-by: Takashi Kajinami <kajinamit@oss.nttdata.com>
Since almalinux-10 label in OpenDev runs on both x86-64-v2 and v3,
and CentOS NFV SIG OVS packages are only build for v3, these jobs
have been only successful when spawned on v3 hardware.
In order to make the job passable on v2 hardware - we should be
building OVS from source.
Rename the jobs to reflect the change
Change-Id: I67b19c18b45af23cda7899f7c2edab21b8ed1ede
Signed-off-by: Michal Nasiadka <mnasiadka@gmail.com>
This change add 4 new varibles to devstack to allow controlling
the behvaior fo downloadign files via get_extra_files
EXTRA_FILES_RETRY=${EXTRA_FILES_RETRY:-3}
EXTRA_FILES_RETRY_ERRORS=${EXTRA_FILES_RETRY_ERRORS:-"500,503"}
EXTRA_FILES_DOWNLOAD_TIMEOUT=${EXTRA_FILES_DOWNLOAD_TIMEOUT:-2}
EXTRA_FILES_RETRY_TIMEOUT=${EXTRA_FILES_RETRY_TIMEOUT:-10}
get_extra_files checks if a file is precached and downlaods
it if not present
The new fucntionaltiy modifes the wget invocation to retry
up to 3 times if there are dns name issues, connection refused
or the server return a 500 or 503
This change will not retry on 403 or 429 error codes
which are often used to indicate rate limits.
The download timeout remains at 2 seconds but
can now be set if required. A 10 second over all retry
wait timeout is added matching wgets default.
finally the use of sed to strip the file:// prefix
if present is replace with usign bash parmater expansion
echo "${file_url#file://}"
Assisted-By: gemini pro 3
Change-Id: Iedbb8974dce34a3d1c89f2a2561df76e696afae6
Signed-off-by: Sean Mooney <work@seanmooney.info>
This was missed in [0].
[0] I6a66359c19d0385beafb4c5e57b6ec3cd6d9cc54
Signed-off-by: Dr. Jens Harbott <harbott@osism.tech>
Change-Id: Idaee45beb63b2badc72c4a70442c5b9aa7fb19fe
The new flag ``TARGET_ENABLE_OVN_AGENT`` will be disabled by default.
If enabled:
* The OVN agent will be enabled, regardless of not being configured.
* The OVN Metadata agent will be disabled, regardless of being
configured.
This variable will be used, initially, in the grenade jobs. It will
be used to test the migration from the OVN Metadata agent to the
OVN agent.
This variable will be removed in 2026.2, along with the OVN Metadata
agent, that is set as deprecated and marked for removal.
Related-Bug: #2112313
Signed-off-by: Rodolfo Alonso Hernandez <ralonsoh@redhat.com>
Change-Id: I8f91e1cb8543da489f495b8cf5196e606a0f5eea
Dropping libapache2-mod-proxy-uwsgi package for debuntu, which is no
longer needed for Jammy, Bookworm and beyond.
libpcre3-dev is removed form the set of packages pre installed
for debian systems.
This change adds both single and two node nodesets for trixie.
Signed-off-by: Dr. Jens Harbott <harbott@osism.tech>
Signed-off-by: Sean Mooney <work@seanmooney.info>
Change-Id: Ib056d2ad64b31657ea8ef9359fed78e589b01b88
Ironic job ironic-tempest-bios-ipmi-direct-tinyipa
has been renamed to ironic-tempest-bios-ipmi-direct
- https://review.opendev.org/c/openstack/ironic/+/950192
Zuul did not give any error or warning when this jobs was
ignored to run and it went unnotice until I manually checked
the gate jobs.
Change-Id: I9b6bb94456ccfd17c2e38cdaa772aef372d169d0
Signed-off-by: Ghanshyam Maan <gmaan@ghanshyammann.com>
Unlike other projects, Keystone was previously enabling validation for
all responses. This is a bad idea (TM). Quoting from the Keystone docs
for the new '[api] response_validation' option added in [1]:
'warn' is the current recommendation for production environments. If
you find it necessary to enable the 'ignore' option, please report the
issues you are seeing to the Keystone team so we can improve our
schemas.
'error' should not be used in a production environment. This is
because schema validation happens *after* the response body has been
generated, meaning any side effects will still happen and the call may
be non-idempotent despite the user receiving a HTTP 500 error.
DevStack is not used for production environments and is instead the
test harness used for the bulk of our integration testing. We *do* want
failed response validation to result in an error, since it will
highlight bugs in our schemas. Thus, we should override the default
value for this option.
[1] https://review.opendev.org/c/openstack/keystone/+/962851
Change-Id: I9fc2c5dce9511165ad2c1ab18db5eb439d357d9b
Signed-off-by: Stephen Finucane <stephenfin@redhat.com>
Related-bug: #2126676
