Age | Commit message (Collapse) | Author |
|
Switch charts / chart to use buffer json
|
|
collectors (#16348)
|
|
* Send node update info only if the host has finished replication
* Log number of hosts replicating / pending to load context
* Remove prefix (thread name is enough)
|
|
Detect database corruption doing query executing and schedule recovery on next restart
|
|
* Update flags during label copy
Update unittests
* Return proper value
* Improve double check in unittests
|
|
|
|
(#16321)
* Keep precompiled statements for alarm log queries to improve performance
* Check bind result
|
|
* split systemd-journal.c
* split fstat caching
* split systemd-journal further
* working systemd-units function
* do not enable systemd-units when libsystemd does not provide the interface
* move the header to the right place
* mixed parantheses
* update codacy exlcusions
* update codacy exlcusions
* update codacy exlcusions
* added option to show show expanded filters by default
* keep the original extension and decode descriptions too
* updated systemd-units function to handle all known unit states
* dont show the path by default
* final touches
* remove trailing spaces
|
|
Add index properly if collision is detected
|
|
Optimize database before shutdown
|
|
* Collectors should not be running at this point, but allow shutdown to continue after several retries (workaround)
* Proceed with shutdown after 10 attempts
|
|
* Prepare metadata sync thread cleanup earlier in the shutdown process
* Set flag for the dimensions that need ML MODEL load instead of queueing a message in the event loop
* Process the dimension ML load during the normal dimension metadata save loop
* Use spinlock for cmd queue / dequeue instead of mutex
Cleanup queue structure
* Remove old ML model load code
* Rebase and cleanup
|
|
Fix label copy, more details in unittest
|
|
* maintain in /tmp/stream-receiver-X.txt a copy of metadata received
* stream log metadata to /tmp/stream-sender-localhost.txt
* log the stream of all senders
* cleanup use of X_update_metadata() functions
* fix for last commit
* rrdlabel unmark/mark/delete unmarked restored
|
|
|
|
|
|
increate metadata version to resend chart upstream when labels and other metadata are updated
|
|
|
|
* cache ctx in collection handle
* cache rd together with rda
* do not repeatedy call rrdcontexts - cached collection status; optimize pluginsd_acquire_dimension()
* fix unit tests
* do the absolutely minimum while updating timestamps, ensure validity during reading them
* when the stream is INTERPOLATED, buffer outstanding data for up to 50ms if the buffer contains DATA only.
* remove the spinlock from mrg
* remove the metric flags that are not used any more
* mrg writers can be different threads
* update first time when latest clean is also updated
* cleanup
* set hot page with a simple atomic operation
* sender sets chart slot for every chart
* work on senders without SLOT
* enable SLOT capability
* send slot at BEGIN when SLOT is enabled
* fix slot generation and parsing
* send slot while re-streaming
* use the sender capabilities, not the receiver
* cleanup
* add slots support to all chart and dimension related plugin commands
* fix condition
* fix calculation
* check sender capabilties
* assign slots in constructors
* we need the dimension slot at the DIMENSION keyword
* more debug info in case of dimension mismatch
* ensure the RRDDIM EXPOSED flag is multi-threaded and set it after the sender buffer has been committed, so that replication will not send dimensions prematurely
* fix renumbering on child restart
* reset rda caching when receiving a chart definition
* optimize pluginsd_end_v2()
* do not do zero sized allocations
* trust the chart slot id of the child
* cleanup charts on pluginsd thread exit
* better cleanup
* find the chart and put it in the slot, if it not already there
* move slots array to host
* initialize pluginsd slots properly
* add slots to replay begin; do not cleanup slots that dont belong to a chart
* cleanup on obsolete
* cleanup slots on obsoletions
* cleanup and renames about obsoletion
* rewrite obsolation service code to remove race conditions
* better service obsoletion log
* added debugging
* more debug
* exposed flag now compares versions
* removed debugging messages
* respolve conflicts
* fix replication check for unsent dimensions
|
|
* move compression header to compression.h
* prototype with zstd compression
* updated capabilities
* no need for resetting compression
* left-over reset function
* use ZSTD_compressStream() instead of ZSTD_compressStream2() for backwards compatibility
* remove call to LZ4_decoderRingBufferSize()
* debug signature failures
* fix the buffers of lz4
* fix decoding of zstd
* detect compression based on initialization; prefer ZSTD over LZ4
* allow both lz4 and zstd
* initialize zstd streams
* define missing ZSTD_CLEVEL_DEFAULT
* log zero compressed size
* debug log
* flush compression buffer
* add sender compression statistics
* removed debugging messages
* do not fail if zstd is not available
* cleanup and buildinfo
* fix max message size, use zstd level 1, add compressio ratio reporting
* use compression level 1
* fix ratio title
* better compression error logs
* for backwards compatibility use buffers of COMPRESSION_MAX_CHUNK
* switch to default compression level
* additional streaming error conditions detection
* do not expose compression stats when compression is not enabled
* test for the right lz4 functions
* moved lz4 and zstd to their own files
* add gzip streaming compression
* gzip error handling
* added unittest for streaming compression
* eliminate a copy of the uncompressed data during zstd compression
* eliminate not needed zstd allocations
* cleanup
* decode gzip with Z_SYNC_FLUSH
* set the decoding gzip algorithm
* user configuration for compression levels and compression algorithms order
* fix exclusion of not preferred compressions
* remove now obsolete compression define, since gzip is always available
* rename compression algorithms order in stream.conf
* move common checks in compression.c
* cleanup
* backwards compatible error checking
|
|
|
|
Fix rrdlabel copy to correctly handle duplicate keys
Enhance the corresponding unit tests
|
|
* Drop unused aclk_alert index
* Log messages only when compiled with NETDATA_INTERNAL_CHECKS
|
|
* Retrieve last connected timestamp from the database (host->last_connected)
* Improve context load performance
Check for agent shutdown while context load in progress
Log information about host load start and finish
* Remove check for slot as it will only reach this part when a slot is found
|
|
Reuse ML load prepared statement and release resources on each batch load
Fix parameter to ML model load to be in seconds not usec
|
|
* Add additional checks
rrdlabels_find_label_with_key_unsafe finds a label with specified key (not only if it exists but with difefrent value)
Quick check if label already exists (avoids JudyLIns)
* Add migration unit test
Add additional unit test to verify that adding a label with the same key will replace its value
|
|
|
|
Use uint64_t for calculations
|
|
Fix log message (queue has no limit now)
Fix unittest
|
|
* dynamic meta queue size
* meta cleanup
|
|
|
|
|
|
* stop the query 250ms before the timeout, to allow sending back partial responses
* on timeout return partial responses
* give it 500ms
* give some additional timeout to plugins.d garbage collection
* define an extension to the timeout for all intermediate hops
* hunting for the crash...
* set value name and len to zero
* remove unneeded memset()
|
|
* Proper init to avoid use after free
* CID 400083 Unchecked return value
|
|
Batch ML model load
|
|
* dont queue removed when there is a newer
* proper concat
* add host_id
|
|
Drop warning when parent is not accepting job status updates
|
|
* Remove unused functions
* No need for prepare statement because the function is not used frequently
* Remove db_meta check, already assumed valid
* Remove D_ACLK_SYNC and D_METADATALOG, fix log message
* Reuse prepared statements per run to avoid sql parsing all the time
* Keep rowid in charts and dimensions
* Host and chart labels keep rowids
* Don't store internal flags
* Remove commented out code
* Formatting
* Fix algorithm when updating dimension
|
|
* remove loading and storing families from alert configs
* remove families from silencers
* remove from alarm log
* start remove from alarm-notify.sh.in
* fix test alarm
* rebase
* remove from api/v1/alarm_log
* remove from alert stream
* remove from config stream
* remove from more
* remove from swagger for health api
* revert md changes
* remove from health cmd api test
|
|
provide a relative_to_absolute function that does not touch the current realtime time
|
|
* For new installation skip database migration steps
* Simplify logging
* Count database tables to determine if database is empty
* Report extended error message
|
|
* Avoid duplicate keys in labels
* Properly delete old label
* Simplify memory statistics logging
|
|
* new aclk schema
* transmit summary to cloud and expose in v2/alerts
* missing assign
|
|
* dyncfg fncnames as constants
* add helper macros to know parser streaming/plugin
* plugins dictionary per RRDHOST
* api_request_v2_config add support for /host/
* streamify pluginsd_register_plugin
* streamify pluginsd_register_module
* streamify report_job_status
* streamify dyncfg get functions
* module_type2str
* add job type and flags
* add DYNCFG_REGISTER_JOB
* implement register job
* push all to parent at startup
* add helper function is_dyncfg_function
* forward virtual functions trough streaming
* separate job2json
* add api/v2/job_statuses
* do cleanup on streaming
* streamify set functions
* support FUNCTION_PAYLOAD trough streaming
* WIP tests
* dont attempt loading non-localhost configs
* move cfg persistence to proper place
* prevent race
* properly update job state at runtime
* cleanup 1
* job2json add missing reason
* add tests
* correct HTTP code
* add test
* streamify delete_job_cb
* add DELETE_JOB keyword
* job delete over streaming
* add tests for create and delete job over parent
* rrdpush common checks to macro
* add missing forwarders
* fix jobs according to test results
* more tests
* review comment 1
* codacy remove valid warning
* codacy ruby fixes
* fix wrong rc check
* minimal test plugin for child
* add test
* dict walk insted of master lock
* minor - english spelling fixes
* thiago comments 1
* minor - rename folder to dynconf
* enable only when built with -DNETDATA_TEST_DYNCFG
* minor - compiler warning
* create dir post daemonization
* stricter URL check
|
|
* Convert a db to WAL with auto vacuum
* Use single sqlite configuration function
* Remove UNUSED statements
|
|
Bind value as 64bit
|
|
* Maintain node's last connected timestamp in the db
* Rebase -- switch to version database v14
|
|
|
|
fix summary field in table
|
|
* add a summary field to alerts
* add summary field to db
* rebase
* better migration
* rebase
* change email notification
* revert to silent
* use macro
* add the summary field to some alerts
* add more summary fields
* change migration function
* add to postgres alerts
* add summary to vernemq
* more summary fields
* more summary fields
* fixes
* add doc
|