Add a new safe_classes configuration option to restrict the classes allowed to be used
as cluster config and event types.
The configuration option allows to specify a comma-separated set of prefixes matched
against the fully qualified class name.
For now, the default value for the configuration is org.graylog.,org.graylog2., which will
allow all classes that Graylog maintains.
This should work out of the box for almost all setups. Changing the default value might
only be necessary if external plugins require cluster config or event types outside the
"org.graylog." or "org.graylog2." namespaces. If that is the case, the configuration setting
can be adjusted to cover this use case, e.b. by setting it to
safe_classes = org.graylog.,org.graylog2.,custom.plugin.namespace.
if said classes are located within the custom.plugin.namespace package.
Refs: https://github.com/Graylog2/graylog2-server/security/advisories/GHSA-p6gg-5hf4-4rgj
* Create Data Stream template, prepare ism policy
* add ism
* create ism and collect jvm heap as first metric
* change lookup and term query to use data streams
* start moving data stream functionality to graylog
* move data stream management into graylog
* code cleanup, add tests
* better validation, clean up after tests
* additional clean up after IT
* don't remove `index-management` plugin from datanode
* inject ObjectMapper
* change number of replicas from default 1 to 0 for test
* set `number_of_replicas` to 0 for .opendistro-ism-config in test
* replace javax with jakarta classes
* fix integration test
* remove data streams on test cleanup
* set high priority for data stream template
* ensure password_secret meets length requirement
* update CL
* update test
* adjusted wording so it references the data node
---------
Co-authored-by: Jan Heise <jan.heise@graylog.com>
* using the hostname as the default node-name, if none was specified explicitly
* using the hostname as the default node-name, if none was specified explicitly
* using the hostname as the default node-name, if none was specified explicitly
* nodeName is either set explicitly in the conf file or will be the result of the hostname setting otherwise
* small refactoring
* refactor datanode directories checks, use configuration validators
* comments, error messages
* fix datanode config dir in docker
* force image pull strategy for datanode dev container
* remove image pull strategy
* More detailed DirectoryWritableValidator for datanode
* remove commons dependency from UploadedCertFilesSecureConfiguration
* basic auth for datanode rest api if security enabled
* added changelog
* adapted to master changes
* configurable rest admin username in datanode
---------
Co-authored-by: Jan Heise <jan.heise@graylog.com>
* jwt implementation, adding preflight + versionprobe
* adding JWT auth to anomaly detection client, some minor refactorings
* added missing license
* added provider, refactorings
* fixing small bugs
* set smaller default expiration
* using the correct OpenSearch distro for the erchitecture
* added config options
* small refactorings, fixing tests, adding dependencies for running tests from inside IntelliJ
* fixing import to correct Duration class
* removing unnecessary debug output
* fixing tests
* fixing tests
* adding JWT to Cluster test
* fixing path
* fixing path problems
* fix architecture logic for DatanodeContainerizedBackend
* fixing path problems
* fixing JWT settings in tests
* fixing JWT settings in tests
* fixing cluster test
* fixing cluster test
* fixing cluster test
* adding missing jwt token
* small refactoring to force thread-safety on creation
* the datanode uses mongodb - so the closing sequence needs to be adjusted
* moving config changes into config file, keep only change that has to be done by code
* adding breaking change info for JWT and clock synchronization.
* renaming config settings
* renaming config settings
* removed debug settings that are no longer in use
* renamed provider/annotation to be more precise
* lowering log level
* smaller refactorings/renamings
* keep config change in code until PR is merged, remove code in a follow-up PR
* moved changelog file to correct folder
* clock synchronsation
Co-authored-by: Dennis Oelkers <dennis@graylog.com>
* added jwt auth into changes after merge
* upped the timeout
* log curl connection information in DatanodeOpensearchWait
* replacing with jwt auth
* changing wait strategy
* fixing imports
* remove uname/pwd from opensearch rest base url
* propagate jwt token provider to http rest client callback, obtain fresh token every time
* inject IndexerJwtAuthTokenProvider in DatanodeConfigurationProvider
* remove rest uname/pwd from datanode configuration
---------
Co-authored-by: Tomas Dvorak <tomas.dvorak@graylog.com>
Co-authored-by: Dennis Oelkers <dennis@graylog.com>
* Insecure startup of OS in datanode only if explicitly configured
* Data node can be started without starting OS, just to be discovered by pre-flight mode
* first version adding gl CA
* adding custom trutstore
* fixed SSLContext
* fixed SSLContext for connectivity check
* fix comments
* fixed imports
* reverting altnames tmp change
* reverting altnames tmp change
* fixed preflightconfig, fixed hostname-verifier
* added credentials to target string, set hostname
* added credentials from target string
* only add TrustManager if https is requested
* fixing tests
* add another alt name
* change creation of restBaseUrl
* DatanodeSecuritySetupIT adaptation for stable hostname of the datanode
* trying to fix naming/binding for docker
* adding another alt name
* added s to http
* changing exception
* more debug output
* change test
* change test
* adding debug
* switch back to http
* more debugging
* more debugging
* added exception to retryer
* fixed boolean eval
* going to the next step on FAILED, too
* fixing if clause
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* cleanup
* removed wrong credentials provider
* added name verifying
* fix alt name handling
* removed unneccessary hostname verifier
* reverted DefaultX509TrustManager
* removed link to DefaultX509TrustManager because it's used in different contexts, singleton for SSL related stuff so we don't instantiate it all the time
* full rollback DefaultX509TrustManager
* reverted change on logging
* revert VersionProbe, unnecessary changes
* DatanodeSecuritySetupIT fixed, separated bind address and hostname
* changed to lambda
* changed to lambda
* small refactorings in csrsigner
* added test for CustomCAX509TrustManager
* added binding for CaService into Preflight
* fixing bindings
* fixing bindings
---------
Co-authored-by: Tomas Dvorak <tomas.dvorak@graylog.com>
* New configuration handling. Incomplete manual config ends with error. Empty config allows to try different config options.
* Trying to reflect functionality responsibilities with better naming
* Change ConfigProvider, so that it stops using PEM files, and starts using PKCS12 files.
Refactoring, splitting the functionality to different classes.
Some config changes that make things work in IDEA.
Reusable truststore.
Truststore stores only root certificates.
* Simpler root certificate finding
* Dir Sync does not remove cert files
* Removed unnecessary default config changes. Cleared confusing certificate copying and unnecessary certificate duplication in 2 folders
* Truststore path created outside of TruststoreCreator. Root certificate found in a different way, without considering its place in the certificate chain
* Fixed log message
* Integration test fixed
* Certs are not an exception in FullDirSync. Introduced second config location as a step towards final configuration. Certificates are moved between locations, as it was done previously
* Fixed a test. Minor refactoring.
* Config location has no default value. It is checked if it is provided, as it is required
* adds a config option to force single-node type so that (at least on macOS) the bootstrap checks won't run and fail
* add changelog
* reworded message
* Refactor data-node download and assembly
- Download OpenSearch aarch64 distribution in addition to x64
- Extract OpenSearch distribution into "target/opensearch/*" instead of
"bin/download".
- Include OpenSearch distributions in "dist/" instead of "bin/" in
assembly
* Fix data-node test container creation
* Adjust OpenSearch location for architecture dependent locations
* If the configured location exists, use it
Otherwise, check for architecture dependent distribution directories.
* Switch default config file to locations that work with the tarball
* Rename datanode.jar to graylog-datanode.jar
* Rename datanodectl to graylog-datanode.sh
* Include graylog-datanode.sh in assembly as bin/graylog-datanode
* Add a jvm.options config file
* Add a production log4j2.xml config file
* Use bundled JVM if it exists
* fixed docker image for datanode integration tests
---------
Co-authored-by: Tomas Dvorak <tomas.dvorak@graylog.com>
