LLview Public Releases
2.4.1-base (November 1, 2025)
Possibility to have tabs on pages (now used for History on production systems, and to be used for Continuous Benchmarks), plus many other internal improvements and fixes.
Added
- Added IO rates to tables;
- Added ENV vars to
$globalvarref; - Added "replay" module (documentation still to be added);
- Added
PRAGMA optimize - Added utility script
dumpconfigto dump YAML config file - Added envvar expansions also for
pre_rowsandrows - Added archiver scripts to compress, tar and move local archived files to remote arch-dirs (documentation still to be added);
- Added 'tabs' key for views
- Added missing
loadmemnodetoLML_DBupdate_file.pm(Should fix GitHub #6) - Added
maskcommaconvert function
Changed
- Improvements on Continuous Benchmark: Added links on names for each benchmark, removed the name column on the benchmark page, made color and style of traces configurable
- Improved documentation, including continuous benchmarks information
- Extended YAML input to allow multiple indices per table
- Improved
waittimeafter a job has started or is in the queue - Changed date format of
DATE_NOW(info_stron JURI) to ISO including timezone
Fixed
- Fixed data collection on Prometheus plugin
- Avoid warnings in db-arch if table is empty
- Fix capitalization for Continuous Benchmark titles
2.4.0-base (June 16, 2025)
New file-parser plugin and extensions to the Prometheus one (more generic for REST-API now). Large rewrite on JuRepTool to generalise plots.
Added
- Slurm plugin: Added query for Slurm accounts from running jobs (still to be added in DB)
- Added YAML-linter
- Added plugin to parse files using regex definitions (e.g., for healthchecker logs)
- Added example configuration for healthchecker (to be used with the file-parser plugin)
- Added example configuration for job error (
joberr.yaml) and node error (nodeerr.yaml) tables - Added new envvar variable
LLVIEW_DEMO_MODEto activate demo mode - JuRepTool: possibility to add red lines to mark graphs (to be used with calibrate)
- JuRepTool: added "Download Data" button for timeline
Changed
- Improved and extended documentation
- Added logo for dark mode in README
- Improved Apache header files
- Changed ActiveSM to percentage
- Extended the 'prometheus' plugin to handle more generic REST-API (possibility to give endpoints, client secret, and more)
- Generalisations on Slurm plugin (unlimited time in queue, format of Gres, empty responses)
- Generalisation on
get_hostnodename.pyto allow multiple expansions - Changed
remoteAll.plandserverAll.plscripts to use same envvars names - Increased timeout on Prometheus plugin
- Improved and optimized workflow on Prometheus plugin (much faster now)
- Changes from production: improved logging, bug fixes, small improvements
- JuRepTool: changed the way the overview graph is defined (now configurable)
- JuRepTool: major rewriting to allow graphs inside the same section get data from different dat files
- JuRepTool: improved CI tests, that should be now faster and include their own configuration
- JuRepTool: added energy values on header, when present on the json file
- Other minor improvements
Fixed
- Fixed
.htaccessfiles to require valid user - Fix for "unparseable" line in slurm output
- Fixed escape sequence for Python>=3.13
- Fixed check of modification date of files, that led to pdf and html reports not being synced.
- JuRepTool: Fixed zoom-lock for new Plotly version
- Other minor fixes
2.3.2-base (December 16, 2024)
Added
- Improved Gitlab plugin (that runs now by default only every 15min), including possibility to give units of metric
- Improvements on Prometheus plugin: possibility to authenticate with token, added min/max to metrics, possibility to turn off verification on requests
- Possibility to give system status information to be shown on the webportal
- Added pre-set options for grid
- JuRepTool: Added 'link failure' error recognition
Changed
- Improved default columns shown on tables (description, conversions, etc)
- Decrease the default amount of cores used in different steps, to avoid using too much memory
- Deactivated all but basic Slurm queries by default (and commented out CB in config)
- Unified 'monitor' logs now also located in the 'logs' folder
- Improved documentation (including a first version of how to add new metrics)
- Changed
onhoverto use list/array instead of dict/object in gitlab plugin (so order is kept) - Adapted
serverAllsearch command to be able to use 2 systems in one server - Changes from production: internal improvements
- JuRepTool: Activated Core metrics by default for JuRepTool reports (must be deactivated if those metrics are not available)
- Other small improvements
Fixed
- Fixes for absent logic cores (for systems without SMT)
- Fixed columns when grid is not used (including defaults)
- Fixed filter for admin jobs on
plotlists.dat(files were not created, but jobs were being added for JuRepTool) - Create temporary
.htgroups_alluser to avoid building up support when there's a problem - Fixes in
monitor_file.pl: folders not recognized in when given with 2 slashes, folders not created when slash at the end missing - JuRepTool: Fixed error output to be also .errlog, to be listed in
listerrors - JuRepTool: Fixed 'CPU Usage' in Overview graph
- JuRepTool: Removed rows containing 'inf' values
- Other small fixes
2.3.1-base (July 10, 2024)
Prometheus plugin and GitLab plugin for Continuous Benchmarks! Many fixes and improvements, some of which are listed below.
Added
- Prometheus and Gitlab (for Continuous Benchmark) plugins
- Brought changes from production version, mainly rsync list of files
- JuRepTool: Added hash for each graph to URL (also automatically while scrolling)
- JuRepTool: Added link in plotly graphs to copy the link
Changed
- Improved README, with thumbnail
- Usage->Utilization for GPU
- Added ActiveSM in GPU metrics
Fixed
- Fixed project link
- Fixed regex pattern for 'CANCELLED by user' to allow more general usernames
- Fix for cases where username is in support but not alluser (previously didn't have access to _queued)
- JuRepTool: Fixed icon sizes in plotly modeBar
- JuRepTool: Fix for horizontal scroll in nav of html report
- JuRepTool: adapt for new slurm 'extern' job name
- JuRepTool: Escape job and step name
- JuRepTool: Ignore '+0' in step id
- JuRepTool: Removed deprecated function 'utcfromtimestamp'
- JuRepTool: Added new tests and fixed old ones (due to new metrics)
- JuRepTool: Added line break in 'Cancelled by username' in PDF timeline to avoid overlapping text
2.3.0-base (May 21, 2024)
Faster tables! Using now ag-grid to virtualise the tables, now many more jobs can be shown on the tables. It also provides a "Quick Filter" (or Global Search) that is applied over all columns at once.
Added
- Support for datatables/grids
- CSV files can be generated
- New template and Perl script to create grid column definitions
- Added
dc-waiqueue on jureptool system config
Changed
- Removed old 'render' field from column definitions (not used)
- Default Support view now has a single 'Jobs' page with running and history jobs using grid
Fixed
- Improved README and Contributing pages
- Fixed text of Light/Dark mode on documentation page
- Fixed get_cmap deprecation in new matplotlib version
2.2.4-base (April 3, 2024)
Added
- Added System tab (usage and statistics) for Support View
- Added option to delete error files on
listerrorsscript - Added
llviewcontroller in scripts (llview stopandllview startfor now) - Added power measurements (
CurrentWatts) (LML, database and JuRepTool) - Added
LLVIEW_WEB_DATAoption on.llview_server_rc(not hardcoded on yaml anymore, as the envvars are expanded forpost_rows) - Added
LLVIEW_WEB_IMAGEoption on.llview_server_rcto change web image file - Added
wserviceandexecdirautomatic folder creation - Added
.llview_server_rcto monitor (otherwise, changes in that file required "hard" restart) - Added
icmapaction, configuration and documentation - Added generation of DBgraphs (from production) to automatically create dependency graphs (shown as mermaid graphs on the "Dependency Graphs" of Support View)
- Added trigger script and step to
dbupdateaction to use on DBs that need triggering - Added options to dump options as JSON or YAML using envvars (
LLMONDB_DUMP_CONFIG_TO_JSONandLLMONDB_DUMP_CONFIG_TO_YAML) - Added
CODE_OF_CONDUCT.md
Changed
- Improved
systemnamein slurm plugin - Changed order on
.llview_server_rcto match.llview_remote_rc - Separated
transferreportsstat step ondbupdate.conf - Moved folder creation msg to log instead of errlog
- Improved documentation about
.htaccessandaccountmap - Improved column group names (now possible with special characters and space)
- Changed name "adapter" to "plugins"
- Improved parsing of envvars (that can now be empty strings) from .conf files
- Further general improvements on texts, logs, error messages and documentation
- JuRepTool: Improvements on documentation and config files
- JuRepTool: Moved config folder outside server folder
Fixed
- Fixed
starttime=unknown - Fixed support in
.htgroupswhen there's no PI/PA - Fixed
'UNLIMITED'time in conversion - Fixed creation of folder on SLURM plugin
- Fixed missing
idon<input>element - Removed export of
.llview_server_rcfrom scripts (as it resulted in errors when in a different location) - JuRepTool: Fixed deprecation messages
2.2.3-base (February 13, 2024)
Added
- Added script to convert account mapping from CSV to XML
- Slurm adapter: Added 'UNKNOWN+MAINTENANCE' state
- Added link to project in Project tab
- Added helper scripts in
$LLVIEW_HOME/scriptsfolder and added this folder in PATH
Changed
- Added more debug information
- Further improved installations instructions
- Slurm adapter: Removed hardcoded way to give system name and added to options in yaml
- Removed error msg from hhmm_short and hhmmss_short, as they can have values that can't be converted (e.g: wall can also have 'UNLIMITED' argument)
- JuRepTool: Changed log file extension
Fixed
- Fixed wall default
- Removed jobs from root and admin also from plotlist.dat (to avoid errors on JuRepTool)
- fixed SQL type for perc_t
- JuRepTool: Fixed loglevel from command line
- JuRepTool: Improved parsing of (key,value) pairs
- JuRepTool: Fixed favicon
- JuRepTool: Fixed timeline zoom sync
- JuRepTool: Removed external js libraries versions
2.2.2-base (January 16, 2024)
Added
- Added link to JURI on README
- Added troubleshooting page on docs
- Added description of step
webserviceon thedbupdateaction - Added timings in Slurm adapter's LML
- Added new queue on JuRepTool
- Possibility to use more than one helper function via
data_pre(from right to left) - Core pattern example configuration (when information of usage per core is available)
Changed
- Changed images on Web Portal to svg
- Improved installations instructions
- Lock PR after merge (CLA action)
- Improved CITATIONS.cff
- Automatically create shareddir in remote Slurm action
- Changed name of crontab logs (to avoid problems in case remote and server run on the same place)
Fixed
- Fixed default values of wall, waittime, timetostart, and rc_wallh
- Improved how logs are cleaned to avoid stuck files
- Fixed workflow of jobs with a single step
2.2.1-base (November 29, 2023)
Changed
- Improved the parsing of values from LML to database
Fixed
- Added missing example configuration files
2.2.0-base (November 13, 2023)
A new package of the new version of LLview was released Open Source on GitHub! Although it does not include all the features of the production version of LLview running internally on the JSC systems, it contains all recent updates of version 2.2.0. On top of that, it was created using a consistent framework collecting all the configurations into few places as possible.
The included features are:
- Slurm adapter (used to collect metrics from Slurm on the system to be monitored)
- The main LLview monitor system that collects and processes the metrics into SQLite3 databases
- JuRepTool, the module to generate HTML and PDF reports
- Example actions and configurations to perform a full workflow of LLview, including:
- collection of metrics
- processing metrics
- compressing and archiving
- transfer of data to Web Server
- presenting metrics to the users
- Jülich Reporting Interface (downloaded separately here), the module to create the portal and present the data to the users
Not included are:
- Client (Live view)
- Other adapters (currently only Slurm)
The documentation page was also updated to include the installation instructions.