Age | Commit message (Collapse) | Author |
|
This PR adds (inactive) support that we will use to fill the gaps on chart when a receiving agent goes offline and the sender reconnects. The streaming component has been reworked to make the connection bi-directional and fix several outstanding bugs in the area.
* Fixed an incorrect case of version negotiation. Removed fatal() on exhaustion of fds.
* Fixed cases that fell through to polling the socket after closing.
* Fixed locking of data related to sender and receiver in the host structure.
* Added fine-grained locks to reduce contention.
* Added circular buffer to sender to prevent starvation in high-latency conditions.
* Fixed case where agent is a proxy and negotiated different streaming versions with sender and receiver.
* Changed interface to new parser to put the buffering code in streaming.
* Fixed the bug that stopped senders from reconnecting after their socket times out - this was part of the scaling fixes that provide an early shortcut path for rejecting connections without lock contention.
* Uses fine-grained locking and a different approach to thread shutdown instead.
* Added liveness detection to connections to allow selection of the best connection.
|
|
Convert the monolithic ebpf.plugin in a modular plugin.
|
|
* Install/Bundle the eBPF kernel-collector library / eBPF programs anyway even if the Kernel check fails (just warn the user)
* Fix missing check-kernel-config.sh from dist (breaking kickstart.sh installs)
* Add missing packaging/ebpf.version and packaging/ebpf.cehcksums to dist source archive
|
|
Implemented a new parser for the pluginsd language
|
|
* Remove old docs generation tooling
* Remove more references to docs/generator/*
* Remove more refernece to things no longer used for docs generation
* Update contributing-documentation.md
* Update contributing-documentation.md
Co-authored-by: Joel Hans <joel.g.hans@gmail.com>
|
|
Brings function used to clean export instances.
|
|
* Add support for spawning processes without pipes.
* Port health_alarm_execute() from mypopen() to netdata_spawn()
* Make alarm notifications asynchronous within a single health thread iteration
* Initial version of spawn server.
* preliminary integration of spawn client with health
|
|
|
|
* Implement formatters
* Add specific configuration options
* Add the connector to the Autotools and CMake configuration
* Initialize a connector instance
* Publish netdata metrics
* Fix internal stats
* Add unit tests
* Improve the documentation
|
|
We renamed eBPF collector for a more meaningful name.
|
|
|
|
This PR merges the feature-branch to make the cloud live. It contains the following work:
Co-authored-by: Andrew Moss <1043609+amoss@users.noreply.github.com(opens in new tab)>
Co-authored-by: Jacek Kolasa <jacek.kolasa@gmail.com(opens in new tab)>
Co-authored-by: Austin S. Hemmelgarn <austin@netdata.cloud(opens in new tab)>
Co-authored-by: James Mills <prologic@shortcircuit.net.au(opens in new tab)>
Co-authored-by: Markos Fountoulakis <44345837+mfundul@users.noreply.github.com(opens in new tab)>
Co-authored-by: Timotej S <6674623+underhood@users.noreply.github.com(opens in new tab)>
Co-authored-by: Stelios Fragkakis <52996999+stelfrag@users.noreply.github.com(opens in new tab)>
* dashboard with new navbars, v1.0-alpha.9: PR #8478
* dashboard v1.0.11: netdata/dashboard#76
Co-authored-by: Jacek Kolasa <jacek.kolasa@gmail.com(opens in new tab)>
* Added installer code to bundle JSON-c if it's not present. PR #8836
Co-authored-by: James Mills <prologic@shortcircuit.net.au(opens in new tab)>
* Fix claiming config PR #8843
* Adds JSON-c as hard dep. for ACLK PR #8838
* Fix SSL renegotiation errors in old versions of openssl. PR #8840. Also - we have a transient problem with opensuse CI so this PR disables them with a commit from @prologic.
Co-authored-by: James Mills <prologic@shortcircuit.net.au(opens in new tab)>
* Fix claiming error handling PR #8850
* Added CI to verify JSON-C bundling code in installer PR #8853
* Make cloud-enabled flag in web/api/v1/info be independent of ACLK build success PR #8866
* Reduce ACLK_STABLE_TIMEOUT from 10 to 3 seconds PR #8871
* remove old-cloud related UI from old dashboard (accessible now via /old suffix) PR #8858
* dashboard v1.0.13 PR #8870
* dashboard v1.0.14 PR #8904
* Provide feedback on proxy setting changes PR #8895
* Change the name of the connect message to update during an ongoing session PR #8927
* Fetch active alarms from alarm_log PR #8944
|
|
* Fix build and add bundle-dashbaord.sh to dist_noinst_DATA
* Fix typo
|
|
This reverts commit e2874320fc027f7ab51ab3e115d5b1889b8fd747.
|
|
|
|
* Add a print function for internal exporting statistics
* Send statistics for simple connectors
* Flush sending buffers on failures
* Send statistics for the Kinesis connector
* Send statistics for the MongoDB connector
* Add unit tests
|
|
* Fix the Prometheus web API code in the exporting engine
* Rename connector types
* Remove the conditional compilation of the exporting engine
* Use labels instead of tags
* Fix the exporter configuration
* Document functions
* Add unit tests
|
|
|
|
Preparing for the cloud release. This changes how we handle the feature flag so that it no longer requires installer switches and can be set from the config file. This still requires internal access to use and is not ready for public access yet.
|
|
* Initial installer components for new dashboard.
* Add script to switch between react and classic dashboards.
* Update to newest version of react dashboard.
* Properly substitute webdir.
* Fix installation of dashboard switch script.
* Properly handle file ownership and permissions.
* Fix install of react dashboard.
* Add dashboard_info.js to the react dashboard tree.
* Update to version 0.3.2 of the React dashboard.
* Switch using file lists instead of nuking old directory.
* Properly handle updates.
* Fix variable naming in switching script.
* Fix copying of files in switching script.
* Fix switching script invocation in installer.
* update dashboard to v0.4.0
* v0.4.1
gauge & easypiechart width update
proper centering when clicking alarm in alarms-log
keep loader until react app is ready
fix top bar not showing when highlighting (few lines of code)
"check known URLs" issue
"reset" options is not hooked up
go to host node from streamed node (low priority, because that's removed on release)
* Persist user selection of dashboard across updates.
* update dashboard to v0.4.2 (temperature, units options fixed)
* v0.4.5
fixed print modal progress text
force reload on sync-selection option change
* fix memory leak on firefox
* fix active-alarms number in main.js header
fix "back to normal" notifications constantly throwing in loop
* fix active-alarms fetching
* support subpaths
Co-authored-by: Jacek Kolasa <jacek.kolasa@gmail.com>
|
|
* Copy files from the MongoDB backend
* Update the documentation
* Rename functions in the MongoDB backend
* Add the connector to the Netdata build
* Add an initializer and a worker
* Add specific configuration options
* Initialize the connector
* Add a ring buffer for inserting data to a MongoDB database
* Add unit tests
|
|
* wip
* add alpn
* beggining of cleanup
* move common code
* add SOCKS5 support
* check HTTP response code
* add timeout
* separate https_client into own files
* fix some mem leaks from master + avoid string copying and alloc/free
* fix some PR unrelated warnings
|
|
* Copy files from the Prometheus remote write backend
* Update the documentation
* Rename backend -> exporting
* Add the connector to the Netdata build
* Separate files for the remote write connector
* Add an initializer and formatters
* Read a connector specific configuration option
* Add a separate function for header sending
* Use labels instead of tags
* Separate write request for every instance
* Add unit tests
|
|
|
|
* Cleaning up the ACLK part 2 (#8187)
* Initial proxy support implementation (#8146)
* Implement the ACLK Challenge-Response Authentication (#8317)
Co-authored-by: Timo <6674623+underhood@users.noreply.github.com>
|
|
|
|
* Fixes for issues not caught at review in #8144. (#8211)
* Properly scrub build environment for external dependencies.
* Remove LWS from our system-level dependencies.
We've decided to just always bundle it in the installer.
* Add missing dependency for Docker build process.
* Use static-build of LWS library created by installer (#8157)
* link static lib of LWS
* only use static LWS by installer
* handle -lcap
* fix problem on debian without lcap
* fix lws check
Co-authored-by: Austin S. Hemmelgarn <ahferroin7@gmail.com>
Co-authored-by: Timo <6674623+underhood@users.noreply.github.com>
|
|
* Prepare files for the AWS Kinesis exporting connector
* Update the documentation
* Rename functions in backends
* Include the connector to the Netdata buid
* Add initializers and a worker
* Add Kinesis specific configuration options
* Add a compile time configuration check
* Remove the connector data structure
* Restore unit tests
* Fix the compile-time configuration check
* Initialize AWS SDK only once
* Don't create an instance for an unknown exporting connector
* Separate client and request outcome data for every instance
* Fix memory cleanup, document functions
* Add unit tests
* Update the documentation
|
|
* syscall_plugin: Compilation
This commit brings the necessaries changes to the compilation files
* syscall_plugin: Collector body
This commit brings the collector body to files.
* syscall_plugin: .gitignore
This commit adds syscall.plugin to .gitignore
* syscall_plugin: Plugin adjust
Fix reference and remove message
* syscall_plugin: Remove limit
Remove call to setrlimit
* syscall: Fix start
This commit fixes problems related with start of the plugin
* syscall_plugin: Bring heartbeat
This commit removes the sleep and changes to heartbeat to avoid plugin receive a SIGTERM
* syscall_plugin: Missing semicolon
* syscall_plugin: Fix dimension
Brings the initial value of chart for the normal dimension of the other values
* syscall_plugin: Fix dimension 2
The previous change did not give the expected results, so I am bringing more a fix
* syscall_plugin: adjust values
Rename function and adjust pid size
* syscall_plugin: Remove Chart and fix var
this commit removes a chart that will not be created and fix an error
when the bytes were calculated
* syscall_plugin: Brings error
This commit brings a new variable that will be used to identify errors
* syscall_plugin: Rename charts
This commit starts to rename the charts properly
* syscall_plugin: Rename plugin
* syscall_plugin: missing changes for rename
* syscall_plugin: fix compilation
* syscall_plugin: bring new charts
* syscall_plugin: Warnings
Remove warnings from compilation time
* vfs_plugin: Fix Error chart plot
There was an error when the chart was being displayed
* vfs_plugin: Change family
This commit changes the family of the VFS plugin
* vfs_plugin: Fix order
This PR fixes the wrong order when creating a chart
* vfs_plugin: Remove path
Remove path from structure
* vfs_plugin: From Perf to HASH
This commit converts the main source a hash table and also split the data collection per chart
* vfs_plugin: Adjusts and exit
This commit brings adjusts to the collect and the complete monitor to exit events
* vfs_plugin: Start process
This commit brings the monitoring of a process start and thread creation to Netdata
* vfs_plugin: Visualization and collection
Adjust variables to show and to collect data
* vfs_plugin: Connection with apps plugin
This commit starts to bring the connection with apps.
* vfs_plugin: Various
This commit brings new label for charts, fix to error chart and adjusts for new charts, I am sorry
* vfs_plugin: basis new chart
This commit brings the basis of the new charts for the plugin
* vfs_plugin: Apps plugin
This commit brings the integration with apps.plugin
* vfs_plugin:fix counter
This commit fixer the difference between apps plugin and counter
* ebpf_plugin: rename charts
This commit renames the charts
* ebpf_plugin: New charts adjusts and log start
* ebpf_plugin: Log thread
Creates the log thread that will be used to store error message
* ebpf_plugin: Rename Web Group
This commit reorganize the charts on dashboard
* ebpf_plugin: Restore
This commit restore the previous status of the collector where we only have a global vision of the problems
* ebpf_plugin: kretprobe
This commit brings the initial changes for the collector works with both eBPF program
* ebpf_plugin: New syscalls
This commit brings the new syscalls that we are monitoring
* ebpf_plugin: New charts
This commit brings new charts to the collector
* ebpf_plugin: Parse config
This commit starts the parser of the file
* ebpf_plugin: collector debug
* ebpf_plugin: Global variables from config
This commit brings the global variable update from the config file
* ebpf_plugin: Clean kprobe_events
This commit brings the clean of kprobe_events and also starts the common library for all eBPF collectors
* ebpf_plugin: Check kernel version
This function brings a check for the kernel version
* ebpf_plugin: Start documentation
This commit brings the initial documentation for the users
* ebpf_plugin: Documentation
This commit brings adjust to code and updates for the documentation
* ebpf_plugin: this commit brings the developer mode to the collector
* ebpf_plugin: Documentation
This commit brings more information to the documentation
* ebpf_plugin: Documentation
This commit brings more information to the documentation
* ebpf_plugin: errno to logs
Brings errno number to logs
* ebpf_plugin: Documentation
This commit brings fixes to the collector documentation
* ebpf_plugin: Move description
This commit move the chart description from the C code to dashboard_info.js
* ebpf_plugin: Rename files
This commit rename files to the final version
* ebpf_plugin: COntinue renaming
This commit continue renaming the files to the final version
* ebpf_plugin: Renaming process
This commit renames the final plugin
* ebpf_plugin: Finish rename
This commit finishes the rename processing
* ebpf_plugin: fix entry charts
This commit removes one chart from mode
* ebpf_plugin: Fix remove
This commit brings a new function to fix the unload of collector when the collector
is running in entry mode
* ebpf_plugin: Rename on old kernels
This commit brings fixes for syscall names
* ebpf_plugin: Timestamp to log
This commit brings the timestamp to the logs
* ebpf_plugin: Remove syscall
With the changes on the backend, we are not monitoring more sys_clone
* ebpf_plugin: The syscall is important for 5.3 or newer, so I am returning
* ebpf_plugin: Remove concurrency
This commit adds variables necessary to interact with the new structor
of the eBPF program
* ebpf_plugin: Ids to dimension
This commit fews the functions name as ids for the dimensions
* ebpf_plugin: Missing chart
This commit brings the missing chart for Netdata
* ebpf_plugin: Remove unecessary message
Remove unecessary error message from the collector
* ebpf_plugin: Rename dimension
This commit renames the dimension for something more meaninful
* ebpf_plugin: Optional log
This commit converts the developer.log in an optional feature
* redirect to stdoou
This commit starts to bring the capability to redirect everything to stdout
* ebpf_plugin: Disable dev mode
This commit removes the possibility to load the dev mode file for while
* ebpf_plugin: Disable eBPF process
By default this plugin won't be enabled
* ebpf_plugin: Update debug message
* ebpf_plugin: this commit adjusts documentation to next release.
* ebpf_plugin: documentation fix.
* ebpf_plugin: Percpu hash
This commit moves from an unique hash table for various to speed up
the collector
* ebpf_plugin: Compatibility
This commit set compatibility version between kernels
|
|
* Fix bundling of libmosquitto for RPM builds.
* Add libmosquitto handling files to make dist output.
|
|
* add aclk_lws_wss_client
* shorten the thread name in case more threads are necessary
* Draft libmosquitto<->libwebsockets integration
* use ringbuffer for recvd data
* Some code cleanup
* if mqtt connection fails close lws connection and reconect
* clear buffers on connection closed
* work on better loop integration
* move mosquitto read out of loop
* remove useless code when using websockets
* LWS - make host and port configurable
* make default port 9002 as we use MQTT over WSS now
* wait for link up before subscribing
start query thread after connection has been made
* cleanup - remove useless var
* if there is anything to write send it immediatelly
* cleanup: move buffers into engine instace
* allow MQTT IO from multiple threads (although preffered is MQTT IO to be done by single thread)
* add warning to future self
* add some comments for whoever reviews
* add destroy fnc - start work on cleanup
* minor - add mosquitto to .gitignore
* fix codacy errors
* do not reconnect automatically by default
* minor - remove outdated comment
* tab -> spaces
Co-Authored-By: Konstantinos Natsakis <5933427+knatsakis@users.noreply.github.com>
* address thiagoftsm valid comments
* add usefull logs in case of trouble
* fix -Wall -Wextra -Wformat-signedness warnings
* log error when connection fails
* update .gitignore to match new installer
* Fwd LWS logs to Netdata logs
* minor - tabulation fixes
* fix comments from thiago
* force SSL
* move UNUSED to libnetdata.h
@thiago correctly pointed out it might be usefull for others
* minor - rename function for clarity
* minor - remove commented out code
Co-authored-by: Konstantinos Natsakis <5933427+knatsakis@users.noreply.github.com>
|
|
* if ACLK enabled require libwebsockets.a
|
|
* - Add initial mqtt support
* [WIP] Agent cloud link
- Setup main mqtt thread to connect to a broker using V5 of the MQTT protocol (TBD)
- Send alarms to "netdata/alarm"
- Add error checks to handle connection failures
- Add params for
Broker, port
Maximum concurrent sent / recev messages
- Dummy function to check claiming status
- Generic mqtt_send command to publish message to a base topic , sub topic
It will end up in the form base_topic/sub_topic
- Add host/port in the connection failure error message
* Test libmosquitto libs
* connect to broker locally (assume localhost:1883)
* subscribe to channel netdata/command
* Test try a reload command to trigger health reload
* publish alerts to netdata/alarm
* - Fix compile issues
* - Use sleep_usec instead of usleep
* - Delay reconnection on failure due to misconfiguration (high cpu usage)
* - Remove the TLS connection config
* - Fix NETDATA_MQTT_INITIALIZATION_SLEEP_WAIT to use seconds
* - Gather ACLK related code under aclk folder
- Add aclk_ functions for abstract layer
- Moved low level libs intergration in mqtt.c
* - Add README.md file with initial comment
* - Clean MQTT v5
* - Code cleanup
* - Remove alarm log for now
- Remove the heart beat
* - Remove message properties for V5
* - Remove message properties for V5 (header)
* Fixed the netdata target to use a local static version of libmosquitto.
The installer does not yet have steps to pull and build the local library.
cd project_root
git clone ssh://git@github.com/netdata/mosquitto mosquitto/
(cd mosquitto/lib && make) # Ignore the cpp error
This will leave mosquitto/lib/libmosquitto.a for the build process to use.
* - Fix compile issues with older < 1.6 libmosquitto lib
* - Enable alarm events to check it works
- Re arrange includes
- Rework topic to be agent/guid/. Actual id will be
returned by the is_agent_claimed
* - Add initial metadata info
- Added helper function in web_api
- Added a debug command (info)
* Update the claiming state to retrieve the claimed id.
* - Use define for constants like command and metadata topics
- Function to wait for initialization of the ACLK link
- New aclk_subscribe command with QOS parameter for the mqtt subscription
- Use the is_agent_claimed function to get the real claim id and use it to build the topics
that will be used for the cloud communication
- Change in netdata-claim.sh.in to write the claim id without a trailing \n
* - Use define for constants like command and metadata topics
- Function to wait for initialization of the ACLK link
- New aclk_subscribe command with QOS parameter for the mqtt subscription
- Use the is_agent_claimed function to get the real claim id and use it to build the topics
that will be used for the cloud communication
- Change in netdata-claim.sh.in to write the claim id without a trailing \n
* - Remove the alarm log for now
- Add code (but disabled) to send charts
* - Use dummy anon, anon as username and password for testing purposes
* - Use client id anon as well
* Testing without TLS
* Switching TLS back on to fix docker environment.
* - Added query processing
An incoming URL now calls web_client_api_request_v1_data to handle a request and push the results
back to the "data" topic
- Move the above processing from the message callback to the query handle loop
- Added helper "pause" , "resume" commands to stop and resume query processing to stress test loading the queue
with queries before executing them
- Changed the endpoint topics to "meta", and "cmd" (previously metadata and command)
* make info message follow protocol
* move metadata msg generation into new func
* move metadata msg generation into new func
* - Add metadata to the responses
- Add hook to queue chart changes on creation and dimensions
- Changed the queue mechanism to include delay for X seconds
- Add delayed submittion of charts to the cloud so that all DIMs are defined to avoid resubmission
* - Add additional data info for aclk_queue command
* - Use web_clinet_api_request_v1 to handle the incoming request
This will handle all requests coming from the cloud
* - Cleanup and aclk_query structure
- Add msg_id parameter
- Enable the incoming JSON request
- Enable the outgoing JSON response
* - Added new thread to handle query processing
- Add lock and cond wait to wakeup thread when queries are submitted
- Cleanup on the main init function
* - Add wait time on agent init, to allow for chart, alarms and other definitions to be completed.
- During the wait time, no queries will be queued
* - Send metadata on query thread init
- New generic create header function for the JSON response
- Pack info and charts into one message
- Modified chart to remove entries (test)
- Modified charts mod to remove entries e.g alarms and volatile info
- Change input to aclk_update_chart (RRDHOST / instead of hostname)
* - When a request fails, add to the payload
- We may need to handle in a different key
- Error check in json parsing
* - Add dummy aclk_update_alarm command
* - Move incoming request JSON parsing code away from mqtt.c
- Added #ifdef ACLK_ENABLE so that we can have code merged but disabled by default
- Added version in incoming and outgoing JSON dict
* - Disable code if ACLK_ENABLE is not defined
- Remove references to the mqtt (mosquitto) lib
- Add dummy stubs in mqtt.c for completeness if ACLK_ENABLE is not defined
* - Disable challenge sample code for now
* - Remove libmosquitto from makefile
* - Fix spaces in Makefile.am
- Remove ifdef to avoid warning from LGTM
* - Remove for now the code that builds an along log test message to send to the cloud
* - Add check for ACLK_ENABLE definition and avoid calling the chart update functions
* - Remove commented code
* - Move source files to the correct place (ACLK_PLUGIN_FILES)
* - Remove include file thats not needed
* - Remove include file thats not needed
- Add improved checks for load_claiming_state()
* - Fix error message. Used error() that also logs errno and message
* - Fix some codacy issues
* - Fix more codacy issues, code cleanup
* - Revert code to address codacy warnings
* - Revert spaces added in a previous commit by mistake
* clean up if/else nest
* print error if fopen fails
* minor - error already logs errno
* - Fix version formatting
* - Cleanup all ACLK related compiler warnings
- Re-arrange include files
- Removed unused defines
* - More compilation warnings fixed
- Bug with thread creation fixed
* - Add condition to skip compilation of the ACLK code entirely. Add env variable ACLK="yes" to enable
* - Add condition to skip the libmosquitto
* - Change feature flag from ACLK_ENABLE to ENABLE_ACLK in accordance with the rest of ENABLE_xx flags
- Typo in info message fix
Co-authored-by: Andrew Moss <1043609+amoss@users.noreply.github.com>
Co-authored-by: Timo <6674623+underhood@users.noreply.github.com>
|
|
|
|
|
|
Initial infrastructure support for agent claiming.
This feature is not currently enabled as we are still finalizing the details of the cloud infrastructure w.r.t. agent claiming. The feature will be enabled when we are ready
to release it.
|
|
* Add top level tests
* Add a skeleton for preparing buffers
* Initialize graphite instance
* Prepare buffers for all instances
* Add Grafite collected value formatter
* Add support for exporting.conf read and parsing
* - Use new exporting_config instead of netdata_config
* Implement Grafite worker
* Disable exporting engine compilation if libuv is not available
* Add mutex locks
- Configure connectors as connector_<type> in sections of exporting.conf
- Change exporting_select_type to check for connector_ fields
* - Override exporting_config structure if there no exporting.conf so that
look ups don't fail and we maintain backwards compatibility
* Separate fixtures in unit tests
* Test exporting_discard_responce
* Test response receiving
* Test buffer sending
* Test simple connector worker
- Instance section has the format connector:instance_name
e.g graphite:my_graphite_instance
- Connectors with : in their name e.g graphite:plaintext are reserved
So graphite:plaintext is not accepted because it would activate an
instance with name "plaintext"
It should be graphite:plaintext:instance_name
* - Enable the add_connector_instance to cleanup the internal structure
by passing NULL,not NULL arguments
* Implement configurable update interval
- Add additional check to verify instance uniqueness across connectors
* Add host and chart filters
* Add the value calculation over a database series
* Add the calculated over stored data graphite connector
* Add tests for graphite connector
* Add JSON connector
* Add tests for JSON formatting functions
* Add OpenTSDB connector
* Add tests for the OpenTSDB connector
* Add temporaty notes to the documentation
|
|
* Checkpoint commit (POC)
* Implemented command server in the daemon
* Add netdatacli implementation
* Added prints in command server setup functions
* Make libuv version 1 a hard dependency for the agent
* Additional documentation
* Improved accuracy of names and documentation
* Fixed documentation
* Fixed buffer overflow
* Added support for exit status in cli. Added prefixes for exit code, stdout and stderr. Fixed parsers.
* Fix compilation errors
* Fix compile errors
* Fix compile errors
* Fix compile error
* Fix linker error for muslc
|
|
* proc.plugin: add pressure stall information
* dashboard_info: add "Pressure" section
* proc.plugin: mention PSI collector in doc
* dashboard_info: fix grammar in PSI section
* proc_pressure: fix wrong line name for "full" metrics
* proc_pressure: fix copypasta
* proc_pressure: refactor to prepare for cgroup changes
* cgroups.plugin: add pressure monitoring
* add proc_pressure.h to targets
* Makefile.am: fix indentation
* cgroups.plugin: remove a useless comment
* cgroups.plugin: fix pressure config name
* proc.plugin: arrange pressure charts under corresponding sections
* dashboard_info: rearrange pressure chart descriptions
* dashboard_info: reword PSI descriptions
|
|
* Start of testing partial requests.
Need to stash this to checkout a PR to test.
* Disambiguated error messages during header validation.
The mocking has blown up in the linker, need to wipe out repo local changes and restart from a known good state.
* Test failures.
CMocka is really not designed for parametric tests which is making it difficult to test the http validation properly.
We have some problems in the web_client.c code that are causing early failures in the testing sequence, and it is
causing CMocka to abort the sequence. Need to try a different approach to building the tests...
* Pedantic style pass.
* Test generation.
There must be another value hidden in the system that CMocka uses. This sets up 3278 tests but the results from
cmocka_run_group_tests_name show 0 tests were run.
* The problem was the "helper"-macro.
Calling CMocka directly, moved the setup/teardown into explicit fixtures. Successfully runs the family of tests over
the same (empty) state.
* Parameterised family of tests runs.
The api_next() acts as a counter, the least significant digit is the prefix_len using the web_buffer in the test_family
struct as a template to walk throufh. The most significant digit is the number of headers to use in the request.
Checked that this walk executes correctly and all the tests run before putting the test payloads back in. We trigger a
failure about 3-4 tests in that takes down the process. Currently investigating which parts are not mocked correctly.
* Pedantic style pass
* Adding a mocking for fatal.
That weird thing with the linker has happened again, need to clean repo and rebuild fresh.
* Full test sequence executes.
The test parameter counter jammed after a failure - we cannot rely on anything in the main test body being executed
after we call the functionailty under test. A failure will skip the rest of the execution.
Moved the counter stepping to the top of the function (i.e. it is now a ++i instead of a i++). Adjusted the initial
state to compensate.
This now steps through all of the test-sequence, but it raises an ugly issue - the post-test cleanup will not be
executed on a failure.
TODO:
* Move the test-state into the test_family.
* Do the clean-up of the previous test (if necesarry) in the step function.
* Fix the assertion on the web_client state.
* Pedantic style pass
* Test state is now in the test_family.
This addresses the issue with leaking on failure and not performing clean-up - we don't really care about memory leaks
during unit-testing, but we do care about reseting the system-under-test back to a known state to guarantee
independence across the tests. The clean-up is now triggered in api_next().
* Flip the wait flag assertions.
Partial requests should leave the web_client waiting to receive more data.
* Fixing ACL flags in test-driver.
This makes some tests pass - but far too many. Probably need a proper debugging function to show the request / response
in a readable format.
* Result from the api mocking.
Setting a successful return code in the api mocking makes the non-partial tests pass.
Zero'ing out the web_client before use has not fixed the initialization errors, there is still some history on the
parse_tries that needs to be tracked down.
Some of the other errors are spurious - they result from stream multiplexing in the testdriver - be careful with less.
* Fix warnings.
Switched the build configuration to CFLAGS="-O1 -ggdb -Wall -Wextra -Wformat-signedness -fstack-protector-all
-DNETDATA_INTERNAL_CHECKS=1 -D_FORTIFY_SOURCE=2 -DNETDATA_VERIFY_LOCKS=1".
The memset introduced last night to zero out the initial web_client state had transposed parameters. Now that the
state is initially zero before hitting the http request processing most issues have disappeared. There are 3000+
passing tests and 48 boundary cases to track down.
* Pushing log entries from each test into a buffer.
This will allow suppression of logs from tests that pass.
* Switched to a unique test definition structure per test.
This cleans up the code as it means that a list of tests can be constructed during the first walk through the parameter
space. There is no need to walk the space twice and keep both walks aligned. Removed the cmocka_unit_test macro and
build the CMUnitTest structures directly -> this allows a real name per test instead of the procedure name.
The walking/step function api_info_next has been folded back into the test procedure as it is simpler to walk the list
in the shared test state.
Current TODO:
* There is a bug, the check on the wait state in the buffer is not being handled properly, investigate why
everything fails.
* The results don't match the old code, are we handing the correct web_buffer to each tested piece of code?
* Capture the test success state -> dump the log buffer on failures.
* State is properly passed through the tests.
Spent a long time chasing a horrible bug that seems to be inside CMocka? The state parameter being passed to each
unit test is different on each call, i.e. it looks like a unique void** where the void * (*state) has been overwritten
with the original value on each iteration through the testing loop. This behaviour does not match the CMocka source
code, which does thread the given valud through the unit test calls. It could be a side-effect of the memory
check-pointing, but the net-effect is that we cannot change the shared state between tests. It can be set in the
setup-fixture and used in each test, but not altered for the subsequent test.
This took a long time to diagnose - the fix is simple, we just share the state in a global pointer. This shared state
is used to walk through the list of test_def structures so that each unit-test knows where it is in the parameter-
space.
* With the correct state the bug in triggering the correct assertions is gone.
* Dump out the buffered logs on test failure.
* The only failing case (relative to these assertions) are the ulta-short partial-requests.
* Check the web_client->mode is set properly.
* Style pass
* Checking values passed to the API despatch point.
* Disabled the parametric tests to do some low-level testings.
Later on both sets of tests will be active. While the low-level url encoding tests are being developed the dynamically
generated set is disabled to make the output easier to read.
Working through the W3C URL spec, against RFC3986 and comparing the cases in available url-parsing test-suites to build
our test-suite.
* Start of the URL test-suite.
The percent-decoding in the current implementation is in the wrong place - it happens too early and causes
non-delimitor characters in the URL to be treated as delimitors. Current unit-tests seem to cover the range of
checks that we need CMocka to make. The handling of output is a little awkward - need something like the dynamic
cases that can output the log on a failure or skip it on a pass.
* Raw material for low-level testing.
* Adding more families in here is getting too messy.
About to switch over to multiple testdrivers.
* Need to clean repo to work around wrapping failures in CMocka.
* CMocka is not compatible with LTO.
The weird wrapping issues that come and go are as a result of LTO. My typical netdata-installer command-line that I use
to reboot the project state disables LTO, while my normal autoreconf / configure command-line does not causing this
problem to reappear seemingly-randomly. To build a single test-driver target this works:
autoreconf -ivf && CFLAGS='-O1 -ggdb -Wall -Wextra -Wformat-signedness -fstack-protector-all -DNETDATA_INTERNAL_CHECKS=1 -D_FORTIFY_SOURCE=2 -DNETDATA_VERIFY_LOCKS=1' ./configure --disable-lto && make web/api/tests/web_api_valid_urls
The actual change in this commit is just a bug-fix.
* Ripping out the parameterized test generator.
Each of the URL cases is slightly and subtly different. This can't be done using the parameterization and will need a
healthy dose of cut and paste.
CMocka does not recognise the mocking for mysendfile, which is necessary to capture the exit route from the URL
parsing.
* Weird bug in CMocka?
For some reason CMocka will not mock out the mysendfile() procedure. We need to mock this to capture the behaviour of
the URL parsing as it is one of the exit paths. The wrapping is setup the same way as for the procedures so I cannot
see any reason that the library would not overwrite the calls. The only difference that I can find is that mysendfile
is in the unit being tested and the other mocked procedures are in different translation units.
This should not make a difference, but we have to disable LTO to get CMocka to work and the symbol patchs is some kind
of linker hack so there could be an issue if LTO is not running and the patch target is inside the same translation
unit.
Hiding it for now with a #ifndef UNIT_TESTING, which then compiles find and control flow hits the mock...
* Converting the ascii comments into unit_tests.
* More nasty cases for unit testing.
The commented out case will trigger a buffer overflow in the netdata agent and crash it.
* Last of the individual unit tests planned before the demo.
* Removing warnings.
* Switching on the rest of the parametric set - the other case with CRs.
* Fix Travis build failure under docker.
* Change the name of a define so it does not collide with existing testing in Travis.
* Add CMocka unit tests to CMake
* Linting pass
* Adding RFC comment to test.
* Buffer overflow checks on the captured logs.
This fixes the seg-fault seen by @vlvkobal and @thiagoftsm during testing.
* Chasing down other valgrind reports.
This gets rid of all of the uninitialised variable warnings. We stil have a memory leak, the headers that are set
during the unit testing switch on compression. This causes the web_client code to call deflatInit2 and allocate
structures for the compressor. We do not have a matching call to deflateEnd anywhere in the code so the memory leaks.
* Cleaning up a comment.
* Fixing review comments from @vlvkobal.
Also noticed that the buffer overflow fix this morning was killing the logfile output, fixed this as well.
* Addressing @thiagoftsm's concerns about the changing number of failures.
Switched the log dump for failing cases to repr().
Found a bug in the test case generator (not storing the flag for `\r`.
Verified that the 58 failing cases are the correct set of failures for the tested code.
|
|
Fixes #7324
|
|
* Use 4 spaces for indentation of non-recipe lines in Makefile.am files
* Be more consistent in the use of space before = in Makefile.am files
|
|
* Start of unit-test for http request processing.
A CMocka test-driver has been added to replay the live-capture headers against the http request processing code. Enough stubs / mocking code have been written to check that we can inject a chosen raw request into the web_client code and check how it was processed. A parameterised buffer building function can setup requests for testing in `web_buffer` structures. Each endpoint in the API will be checked against a variety of http requests. This PoC demonstrates enough functionality to show building the request on demand for each test-case and cleaning up afterwards.
|
|
* [proc.plugin/proc_pagetypeinfo] Initial commit
* [Fix] Generate graphs for pagetypeinfo
* [Fix] Create node/zone/type graphs
* [Fix] Use directly size and order
* [Add] Configuration handling
* [Imp] Changed SetId to identify NodeNumber
* [Fix] Standard name for chart priority and value
* [Fix] use dynamic pagesize
* [Enh] allow prefix for containerized netdata
* [Fix] global system graph always on, but for explicit no
* [Fix] Add more checks for pageorders_cnt and really use it
* [Enh] Special config value of netdata_zero_metrics_enabled
* [Fix] Check we parsed at least a valid line
|
|
|