This is the multi-page printable view of this section. Click here to print.
Blog
Releases
Release cc-backend v1.4.0
Changelog
New Features
- feat: Add buffered channel with worker thread for job start API
- feat: Add tag scopes to front and backend, initial commit
- feat: Add total energy and energy footprint
- feat: SyncUserOnLogin now updates name of token logged user
- feat: add configurability to frontend plot zoom
- feat: add dropdown to user and project list navbar
- feat: add edit of notice box content to admin settings
- feat: add updateUserOnLogin config option for oidc, jwt
- feat: change statistics render of metric plot to min/max/median
- feat: change to resolution increase on zoom
- feat: display energy usage in job view
- feat: make cron worker frequency configurable
- feat: make quick select starttimes url copyable
- feat: move tag management to new job view header
- feat: redirect to requested page after login, solves #281
- feat: split concurrent jobs list to own scrollable component
Bug fixes
- fix: Replace reserved keywords in database schemas
- fix: SimpleResampler fixed
- fix: Update to resampler handling different resolutions
- fix: ad dmissing resampleConfig handling to scope select
- fix: add accelerator scope to to-be archived scopes
- fix: add additionally loaded scopes to statsTable again
- fix: add compatibility for footprint metrics without config
- fix: add missing default resolution case
- fix: add resolution 60 default to ccms nodeData query
- fix: archived statisticsSeries with mean data now shown again
- fix: errors in import paths
- fix: fix api test router init
- fix: fix crashing job view if roofline metrics missing
- fix: fix db migration to v8, changes key name to cpu_load
- fix: fix footprint logic, do not scale thresholds on multi node jobs
- fix: fix getMetricConfigDeep util function
- fix: fix job list render for continuous mode on filter or sort changes
- fix: fix plot labeling if specific host selected, hide loadall if only node returned
- fix: fix plot render for summed metrics on scope change
- fix: fix svelte source paths in makefile
- fix: fix tag filter results
- fix: fix wrongly inserted gql request and import path error
- fix: fixed and changed to footprint update by transactions
- fix: omit resources prop from metricPlot, use series for legend instead
- fix: oversight error on redirect target
- fix: prevent addition of existing scopes to table
- fix: prevent jump to table head on continuous scroll load
- fix: setup user in api test config
- fix: solve inconsistencies with filters, fixes #280
- fix: use configured footprint statType for update
- fix: use left join to keep unmatched stats query result rows
- fix: user name join not required for normal jobStats
- fix: wrong display of tag after filter select
Release notes
This is a minor release of cc-backend
, the API backend and frontend
implementation of ClusterCockpit.
Supports job archive version 2 and database version 8. Please check out the Release Notes for further details on breaking changes.
Breaking changes
- You need to perform a database migration. Depending on your database size the migration might require several hours!
- You need to adapt the
cluster.json
configuration files in the job-archive, add new required attributes to the metric list and after that edit./job-archive/version.txt
to version 2. - Continuous scrolling is default now in all job lists. You can change this back to paging globally, also every user can configure to use paging or continuous scrolling individually.
- Tags have a scope now. Existing tags will get global scope in the database migration.
New features
- Tags have a scope now. Tags created by a basic user are only visible by that user. Tags created by an admin/support role can be configured to be visible by all users (global scope) or only be admin/support role.
- Re-sampling support for running (requires a recent
cc-metric-store
) and archived jobs. This greatly speeds up loading of large or very long jobs. You need to add the new configuration keyenable-resampling
to theconfig.json
file. - For finished jobs a total job energy is shown in the job view.
- Continuous scrolling in job lists is default now.
- All database queries (especially for sqlite) were optimized resulting in dramatically faster load times.
- A performance and energy footprint can be freely configured on a per subcluster base. One can filter for footprint statistics for running and finished jobs.
Known issues
- Currently energy footprint metrics of type energy are ignored for calculating total energy.
- Resampling for running jobs only works with cc-metric-store
- With energy footprint metrics of type power the unit is ignored and it is assumed the metric has the unit Watt.
Download
Release cc-backend v1.3.1
Changelog
New Features
- ef51e69 feat: Add roofline color scale for time information
- 5757530 feat: add jobname filter to joblist textfilter
- 72557fd feat: add statistics series render to job view metric plots
- 8d1228c feat: rework list searchbar, adds project-specific mode, add to user-joblist
Bug fixes
- 54f7980 fix: Add required key to init config file
- 597bccc fix: add SQL JSON validity check to meta_data query
- 320c87a fix: add additional 30d fitler to searchbar fallback handling
- a4397d5 fix: add scramble to textfilter component
- 70e6376 fix: allow single partial errors on otherwise non-empty returned metric array
- ba1658b fix: correct selectable histogram placement in status view
- b48d1b8 fix: correct status view columns on mobile displays
- 061c9f0 fix: deselected metrics were marked as missing on new jobview load
- 420bec7 fix: fix jobname and arrayjobid timeouts by adding additional 30d filter
- c9eb40f fix: fix metricPlot y zoom reset
- cbaeffd fix: improve speed of hasNextPage query for infinite scroll
- 4344c26 fix: make foorprint from statsSeries nullsafe
Supports job archive version 1 and database version 6. Please check out the Release Notes for further details on breaking changes.
Release notes
This is a bugfix release of cc-backend
, the API backend and frontend
implementation of ClusterCockpit.
Breaking changes
None
Download
Release cc-backend v1.3.0
Changelog
New Features
- feat: Add OpenID Connect Authentication support
- feat: Add cluster config endpoint to rest api
- feat: Add jobs endpoint to retrieve job meta and all job metric data
- feat: Add rest endpoint to add/edit Metadata entry
- feat: Allow to revert db to previous version
- feat: add footprint card displaying basic metrics
- feat: add selectable histograms to status view
- feat: prototype infinite scroll implementation
Security updates
- sec: update dependencies
Bug fixes
- fix: Adapt tag db queries to also work with mysql/mariadb
- fix: Use peak threshold for render limit maxy
- fix: add acc scope to job query if acc >= 1
- fix: fix scope autoselect on jobview statstable
- fix: make hasnextpage optional parameter, use only if inf scroll configured
- fix: move scroll event behind condition
- fix: multiple accs with identical label, cloned data for single acc
- fix: retrigger gql api at manual refresh
- fix: trigger continuous load condition earlier
Supports job archive version 1 and database version 7.
Release notes
This is a minor release of cc-backend
, the API backend and frontend
implementation of ClusterCockpit.
Breaking changes
This release fixes bugs in the MySQL/MariaDB database schema. For this reason
you have to migrate your database using the -migrate-db
switch.
Download
Release cc-backend v1.2.2
Changelog
Bug fixes
- 4b06fa7 fix: Fix buggy logic and simplify code if ValidateUser enabled
- bb1c8cc fix: Move name extract from token in else branch
- e61ff01 fix: adapt roofline render to browser zoomlevel Supports job archive version 1 and database version 6.
Release notes
This is a bugfix release of cc-backend
, the API backend and frontend
implementation of ClusterCockpit.
** Breaking changes **
The LDAP configuration option
user_filter
was changed and now should not include the uid wildcard. Example:- Old:
"user_filter": "(&(objectclass=posixAccount)(uid=*))"
- New:
"user_filter": "(&(objectclass=posixAccount))"
- Old:
The aggregate job statistic core hours is now computed using the job table column
num_hwthreads
. In a future release this column will be renamed tonum_cores
. For correct display of core hoursnum_hwthreads
must be correctly filled on job start. If your existing jobs do not provide the correct value in this column then you can set this with one SQL INSERT statement. This only applies if you have exclusive jobs, only. Please be aware that we treat this column as it is the number of cores. In case you have SMT enabled andnum_hwthreads
is not the number of cores the core hours will be too high by a factor!The jwts key is now mandatory in config.json. It has to set max-age for validity. Some key names have changed, please refer to config documentation for details.
The following API endpoints are only accessible from IPs registered using the apiAllowedIPs configuration option:
/users/
[GET, POST, DELETE]/user/{id}
[POST]
** NOTE **
If you are using the sqlite3 backend the PRAGMA
option foreign_keys
must be
explicitly set to ON. If using the sqlite3 console it is per default set to
OFF! On every console session you must set:
sqlite> PRAGMA foreign_keys = ON;
Otherwise if you delete jobs the jobtag relation table will not be updated accordingly!
Release cc-backend v1.2.0
Changelog
New Features
- 742c2e3 feat: Add uplot histogram, implemented in userview
- f6c4c96 feat: Add users rest endpoint swagger docs
- da8cefe feat: Change histogram to piechart in status view
- c4a9fcc feat: Implemented rooflineplot with uPlot
- e80ce7a feat: Rework analysis view top to contain piechart
- b42a11d feat: Use chart.js for polarplot n jobview
- 36abed2 feat: add auto-reloading to system and node views
- 423e800 feat: add hover-legend to histograms & metricplots
- 3014f59 feat: add new distribution plots to status view
- 59c749a feat: add select to analysis view pie chart
- f933cad feat: add select to status view pie charts
- ed056b0 feat: add sorting in sub-node scopes in statsTable
- 2655bda feat: enable uplot XY-Zoom for metrics
- b623092 feat: persist analysis and status pie selections
Bug fixes
- 80be786 fix: Responsive Navbar
- 6a1e351 fix: analysis metric histogram normalized by scope
- 4eceab4 fix: change analysis top users to core hours
- 8a473de fix: core/accelerator scope in statstable on load
- 4244a37 fix: correct timestamp logic in node-view
- f286872 fix: hover legend display now depends on datasize
- 69ee19b fix: include running jobs case in statsQueries
- 129e6a6 fix: metric y-range render limit for data outliers
- c84a0fb fix: push bootstrap to v5.3.1 and icons to v1.10.5
Supports job archive version 1 and database version 6.
Release cc-backend v1.1.0
Changelog
New Features
- c6a0d44 feat: Add optional notification banner on homepage
Bug fixes
- eed8bb2 fix: add scrambling to user names and projectIds
- 58c7b0d fix: check if acc metrics are acc scope by default
Supports job archive version 1 and database version 6.
Release cc-backend v1.0.0
We are happy to announce the initial stable release of cc-backend, the web- and api-backend of ClusterCockpit.
Supports job archive version 1 and database version 6.
News
Talk at deRSE24 Conference for Research Software Engineering
Jan Eitzinger will give a talk on the development of ClusterCockpit at the Conference for Research Software Engineering (deRSE24) . The deRSE24 conference takes place from 5 Mar 2024 to 7 Mar 2024 in Wuerzburg, Germany.
See you there!