| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
This commit marks the release of sos-4.5.1
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
| |
Signed-off-by: Arif Ali <arif.ali@canonical.com>
|
|
|
|
|
|
|
|
|
|
| |
Collect "ostree fsck" via the plugin specific option, not via --verify
option.
Relevant: rhbz#2161533
Resolves: #3147
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
| |
There is no need to have that option enabled by default for RHOCP.
Relevant: #3147
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
| |
/etc/osbuild-composer contains customizations users create
for Image Builder. This directory is not currently owned by
any package but if present, it, and all files inside,
should be gathered as well.
Related: RHBZ#2169776
Signed-off-by: Jose Castillo <jcastillo@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
sos report should print generated sosreport directory or tarball or
etimate of size even with --quiet option.
Resolves: #3142
Closes: #3143
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
| |
Signed-off-by: Pablo Acevedo Montserrat <pacevedo@redhat.com>
|
|
|
|
| |
Signed-off-by: Pablo Acevedo Montserrat <pacevedo@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
registry_passwords occur in rhui-tools.conf (#3138) but also in (backups
of) answers.yaml files.
Relevant: #3138
Resolves: #3140
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
| |
Scrub registry_password value in /etc/rhui/rhui-tools.conf
Relevant: rhbz#2170893
Resolves: #3138
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Commits a downstream test from RHEL to ensure that a given release has a
consistent version string in the UI report, manifest, and most
importantly for the package nvr for an sos package built from the
current branch. Note that this test is only implemented for RPM
installations at this time.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
| |
Add test for #3071 : Prevent obfuscating tmpDir
Resolves: #3136
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
| |
Add unit test for #3030.
Relevant: #3030
Resolves: #3135
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Along with existing Ceph plugins, Ceph iSCSI gateway use case requires
additional config and log files to be analyzed.
https://docs.ceph.com/en/latest/rbd/iscsi-overview/
Closes: #3098
Related-Bug: https://bugs.launchpad.net/ubuntu/+source/sosreport/+bug/2000672
Signed-off-by: Nobuto Murata <nobuto.murata@canonical.com>
|
|
|
|
|
|
|
|
| |
- add `--noproxy` option
- update api-version to `2021-01-01`
- add query parameter `format=json`
Signed-off-by: Toshi Aoyama <toaoyama@microsoft.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Every output of a command run on a container is saved
in the sos_commands folder, also additionally a symbolic
link is created from the sos_container folder to link the
container where it was run to the output file. When there
is a duplicated command on the list to be run on the
container, the symbolic link creation will fail due to
file already exists error.
To avoid this error we can trust on the check_path method,
that will give us a clue avoid if the file already exists or
not to continue with the symbolic link creation.
Signed-off-by: Fernando Royo <froyo@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When the snap is being packaged via core22, it used python3.10
but bionic used python3.6 and seems have runtime issues there.
Added the sos alias for testing, to ensure that when the stage
one testing is done, that sos is correctly tested within the
snap
This alternative ensures that the packaging module is packaged
in the snap, and can use this if the pkg_resources version does
not work
Add a wrapper function in utlities, so that the parse_version
can be used in a consistent way across the sos project
Closes: #3126
Signed-off-by: Arif Ali <arif.ali@canonical.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
When a list of command are sent to run on a container the output
is printed to file on sos_commands/<plugin> folder, but also a
symbolic link is created from the sos_container/<container> file
to the sos_commands/<plugin> file. When a command is duplicated,
we get an exception on the plugin and the rest of commands are not
executed.
This patch removes some hardcoded commands, and clean the list from
duplicated commands before sending to the container.
|
|
|
|
|
|
|
|
|
| |
* Added missing tags for plugins to match insights spec names, so they
can be queried by insights when loading an archive. This will aid in
spec loading via the manifest instead of defining the specs
individually as we do now.
Signed-off-by: Ryan Blakley <rblakley@redhat.com>
|
|
|
|
|
|
| |
This commit marks the release of sos-4.5.0.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
| |
Updates the log patterns for `ceph_common` to handle Ceph versions that
use or don't use fsid's as part of the logging paths.
Related: #3122
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Simplify the log collection condition by introducing a file glob. It can
capture logs from both existing clusters such as package based
deployments and container based deployments such as cephadm.
Closes: #3100
Co-authored-by: Samuel Walladge <samuel.walladge@canonical.com>
Signed-off-by: Nobuto Murata <nobuto.murata@canonical.com>
|
|
|
|
|
|
|
|
|
| |
When no case id is provided during upload, a prompt of username must fail over
to SFTP upload.
Relevant: #3117
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
| |
Since prompting for upload username/password relies on case_id for some
policies, let fully set case_id before that.
Resolves: #3117
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
| |
Relates: #3060
Signed-off-by: Miroslav Hradilek <mhradile@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Fix a corner case in separating host address from an IPv6 address when
the host string is (substring of) the network address string.
Resolves: #3121
Relevant: #3120
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit allows `Policy` to inherit the logging streams used by the
rest of sos. With this commit, this is primarily the `ui_log` in order
to report info and error level messages to the console. As was
discovered earlier, avocado and potentially other automation may
sometimes miss `print()` output, and so we should consistently use the
logging mechanism.
Future commits may be expected to enhace debug logging of Policy
functionality.
See also: #3106
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
| |
Signed-off-by: Trevor Benson <trevor.benson@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add all the new versions since 7, and update name/versions based on
https://wiki.debian.org/DebianReleases. Checked latest bookworm and
the release string was in /etc/debian_version, but
/etc/debian_version was inconsistent on other versions.
/etc/lsb-release didnt exist by default.
grab the codename from /etc/os-release and base the versions from
there
forky and trixie and not out yet, but added as per the releases page
for future proofing.
Closes: #2691
Signed-off-by: Arif Ali <arif.ali@canonical.com>
|
|
|
|
|
|
|
|
|
| |
Collect /etc/foreman_scap_client/config.yaml on any Foreman OpenSCAP
client system.
Resolves: #3115
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Update most of the GCE images we use for CI testing to their latest
releases for their respective versions.
This commit also replaces F35 testing with F37 beta. F35 is now EOL and
as such no future sos releases would be included there.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Switch to using `parse_version()` from setuptools instead of
`LooseVersion()` from distutils, since distutils is being removed in
python 3.12.
Related: #3093
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In python 3.12 distutils will be removed. As such, we need to update to
the replacement `setuptools`.
This commit makes the basic change over in `setup.py`, so that an
`sdist` source tarball can be generated. Note that while this source
tarball will still have the `.po` files in it any build tarball
(`bdist`) produced via the new `setup.py` will *not* have `.mo`
translation files compiled and included at this point.
In reviewing this change, it was found that our internationalization is
currently broken and very out of date. Future work will focus on fixing
that situation, but for now the immediate packaging needs are being
addressed.
Resolves: #3093
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
| |
Enhance on timeout mechanism. Force the parent process to exit when
it has a deadlock child.
Signed-off-by: Junius Gao <Junius.Gao@veritas.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, mocked files were kept under the `tests/test_data/`
directory and generally mimic'd the file location they would be
temporarily copied to during the execution of their relevant tests.
This has a few maintainability drawbacks, and the handling of the
`files` attribute for test cases as either strings or tuples is at best
confusing.
Improve on this by instead making the `files` references relative to
where the test case file is written. This enables easier maintenance by
keeping all test requirements closer together, rather than spread across
the repo. As such, the `files` attribute now requires a list of tuples,
taking the form `(relative_src, absolute_dest)`. Additionally, fake
plugins for tests that need them to artificially test a specific
criteria should also be included in the test's subdir now.
Along with this change, move several StageTwo tests to their own subdirs
that now contain both the test cases and the needed files for mocking.
This should be the new design pattern going forward - if a test needs to
mock files of any kind, put it in a new subdirectory (and if it doesn't
need to mock files, continue to keep it in the relevant directory within
the test suite).
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
| |
Signed-off-by: Nikhil Kshirsagar <nikhil.kshirsagar@canonical.com>
|
|
|
|
|
|
| |
Resolves: #3109
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
It was recently found that some condition will cause Avocado to not
capture trailing `print()` statements in our test suite, and it is
reasonable to assume other automation may also have similar edge cases.
Resolve this by switching potentially problematic `print()`s to use the
ui logging stream, which will still print to console even after the file
handler has been closed.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
| |
Signed-off-by: zoedong <zoedong@tencent.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Previously, remote command executions handled by policies were done by
moodifying the command string based on the `remote_exec` property of the
given `SoSTransport` in use for the node that the policy was loaded for.
While this worked well for SSH connections, newer transports may need to
do some manipulation of returned data in order for the rest of `sos
collect` to function as intended.
As such, switch to directly using a transport's `run_command()` method,
which will ideally handle any needed manipulations of either how the
command is execute and/or how the returned data is presented to the
calling component.
Related: #3087
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
| |
The string or list strings in the param `copyspecs`
are processed as globs, not regular expressions.
Signed-off-by: Samuel Walladge <samuel.walladge@canonical.com>
|
|
|
|
| |
Signed-off-by: lilinjie <lilinjie@uniontech.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The changes in respecting strict_node_list are three-fold:
1) Don't add local hostname among "list of nodes to collect from:"
2) Skip explicit adding of the primary node to client_list
3) Apply strict_node_list to reduce_node_list (as it can purge away
hostname or IP address of the local host, otherwise)
Resolves: #3096
Signed-off-by: Pavel Moravec <pmoravec@redhat.com>
|
|
|
|
|
|
|
| |
Updates the plugin to account for newer versions of Ceph, similar to the
previous few commits focusing on the ceph plugins.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Newer versions of Ceph, such as for RHCS 5, make some changes to the
container names as well as how to leverage `ceph daemon` commands.
Update the plugin to reflect this, and use the available admin sockets
on the host to capture that output, rather than always running it inside
the osd container.
Further, account for the use of an fsid in directory paths for newer ceph.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Newer versions of ceph do not have running processes directly on the
node. As such, the current method of determining the ID for the `ceph
tell` commands is not functional. Fix this by using the output of `ceph
status` on RHCS 5 nodes to get the node IDs.
Further, while ceph containers can be used to enable this plugin, those
containers by default cannot run various `ceph` commands collected by
the plugin. However, those commands are functional directly on the host,
so no longer attempt to execute within the containers.
Finally, update `postproc()` for newer versions as well.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous changes that broke up the ceph plugin into smaller
component specific plugins inadvertently broke the plugin enablement by
overriding the `check_enabled()` function to check for file presence
under a directory. This in turn replaced the standard checks, such as
the presence of certain containers.
Fix this by removing the method override, and leveraging the `files`
tuple against the component-specific directory into which the globs were
trying to check.
Further, update the container name regexes for enablement as they have
changed slightly since the initial plugin creation.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
Adds a new cluster profile for RHCS 5 for `sos collect`. This profile
depends upon the use of `cephadm` which is used to both deploy and
manage the cluster. Users may optionally restrict the list of nodes to
collect from by using the `-c ceph.labels` option to specify a set of
label(s) to filter node results with.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Clusters can already pass plugin options to nodes, and can also set
arbitrary options on individual nodes or primary nodes. So, rather than
requiring a cluster profile specify sos options to both nodes and
primaries, instead allow profiles to specify options via a simple
`sos_options` dict that will get applied to every node automatically, as
is the case with plugin options.
Note that user values for these options will override cluster values.
For example, if a cluster spceifies a `--log-size` value, and the user
does on the command line, then the user's value will have precedence.
Signed-off-by: Jake Hunsaker <jhunsake@redhat.com>
|
|
|
|
|
|
| |
Circle Linux is an production-ready distribution from RHEL.
Signed-off-by: Bella Zhang <bella@cclinux.org>
|